9410f7e6e6
* test(#3897): failing-first coverage for §8.3 rungs 2-4 ADR-3473 §8.3 has four rungs; #3883/PR #3896 shipped the first. This pins the other three RED before any fix. Rung 2 — the install marker has four readers and resolveRuntime is not one. resolveRuntime resolves GSD_RUNTIME > config.runtime > 'claude' and reads no marker at all, while bin/install.js writes one (#2297) and FOUR hand-rolled readInstallRuntimeMarker copies exist: src/model-resolver.cts:65 (cached, with test seams), hooks/gsd-agent-isolation-guard.js:112, and TWICE in hooks/gsd-cursor-subagent-start.js at :346 and :355. Four copies of one rule. Fixtures and seam names mined from PR #3382 rather than re-derived; it implemented this rung and was closed "not on the merits". Rung 3 — the sandbox map, and the fallback that was the real defect. Measured across all 35 files in agents/, deriving workspace-write iff tools: declares Write or Edit: - all 11 CODEX_AGENT_SANDBOX entries derive to their mapped value exactly, zero disagreements — the map carries nothing the contract does not - 24 roles fall through `|| 'read-only'`, of which 16 declare Write or Edit So the map is redundant and the silent fallback is the defect. The maintainer chose to derive but hold those 16 at read-only pending the question of whether Codex enforces sandbox_mode or merely advises; HALT.md records it. T20 asserts the emitted sandbox_mode PER ROLE against a captured baseline, not in aggregate — an aggregate passes while one role silently widens, which is the proxy-instead-of-identity shape this repo names. T24 and T25 fail on a stale hold, so the hold list cannot rot into the subset map being deleted. Rung 4 — shortFormToId, recovered rather than invented. I nearly reported this as another wrong §8.3 claim: `git log -S shortFormToId` returns only documentation commits. That was the wrong instrument. Direct inspection of sdk/src/query/phase.ts at 11918dcc3^ shows five occurrences, and the tests match that code rather than a guess at its semantics — including first-write-wins on a duplicate short form. T43 asserts at the consumer's output: the emitted `waves` map from the real CLI, which pre-fix collapses to {"1":[...]} because every short-form edge is dropped. A unit assertion on resolveDependencyId would have passed throughout this defect's life. Observed RED, this tree: rung 2 11/11 fail — no marker rung, no seams rung 3 T23,T24,T25,T26,T30 fail; T28 fails (validate agents passes a TOML whose sandbox_mode disagrees — it checks presence only) rung 4 T42,T44 fail; T43,T49 fail with waves collapsed to a single wave 1 Green and staying green: T20/T21/T22/T27 as captured baselines, #3885's unresolvable-token warning and wave-verdict suppression, and #3785's display-mapping passthrough. If the third tier over-reaches, those go red — that is their job. Disclosed weakness: T45 (a canonical id with no dash is not short-form indexed) cannot be isolated behaviorally, because planMap always masks it. It is a non-crash boundary pin, weaker than the other rows, and is recorded as such rather than presented as equivalent. Design: .gsd/phase/feat-3897-adr3473-83-rungs/40-design.md Test matrix: .gsd/phase/feat-3897-adr3473-83-rungs/50-test-matrix.md Decision: .gsd/phase/feat-3897-adr3473-83-rungs/HALT.md Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enhance(#3897): §8.3 rungs 2-4 — one marker reader, a derived sandbox, the third depends_on tier ADR-3473 §8.3 has four rungs. #3883/PR #3896 shipped the first. These are the other three. Rung 2 — the install marker had four readers, and resolveRuntime was not one. resolveRuntime resolved GSD_RUNTIME > config.runtime > 'claude' and read no marker, while bin/install.js writes one (#2297) and four hand-rolled readInstallRuntimeMarker copies existed: src/model-resolver.cts (cached, with seams), hooks/gsd-agent-isolation-guard.js, and twice in hooks/gsd-cursor-subagent-start.js. model-resolver's was already the house idiom, so it was promoted rather than replaced: src/runtime-slash.cts now owns it, and model-resolver plus both hooks delegate. The hooks reach it through ensureRuntimeBuild(), the seam lint-hooks-runtime-build-seam enforces. No import cycle existed - checked both directions before moving anything. The marker is the THIRD rung: env > project config > marker > 'claude'. N1 was checked rather than assumed, and my first reading of it was wrong. A marker holding an unknown name comes back essentially verbatim, which looked like a validation gap. Measured against the env rung with the same inputs - including "../../etc/passwd" and "claude;rm -rf /" - the two are identical, because they share resolveRuntimeNameFromCandidates. N1 asks for exactly that, and it is met. The residual (the shared normalizer normalizes shape, it does not validate against the known-runtime set) is pre-existing on the env rung and plausibly deliberate, since a new runtime should not need a code change. The marker also does not widen the trust boundary in any real sense: it lives inside the install tree beside the code, so anyone who can write it can write runtime-slash.cjs itself. Rung 3 — the map was redundant; the silent fallback was the defect. Measured across all 35 files in agents/, deriving workspace-write iff tools: declares Write or Edit: all 11 CODEX_AGENT_SANDBOX entries derive to their mapped value exactly, zero disagreements. The map carried nothing the contract did not already have, so it is DELETED rather than reconciled. What was actually broken is `|| 'read-only'`, which silently under-granted 24 of 35 roles. 16 of those 24 declare Write or Edit and would widen under derivation. Per the maintainer's decision (HALT.md), they are held at read-only pending the question of whether Codex enforces sandbox_mode or merely advises. Emitted TOML is therefore byte-identical for all 35 roles - asserted per role, not in aggregate, because an aggregate passes while one role silently widens. The hold list self-invalidates. A hold whose role no longer derives broader fails, and so does a hold naming a role with no agents/<name>.md. Without that it would rot into exactly the hand-maintained subset map being deleted, and this commit's own ledger claim would become false over time. Both cases were proved by injecting them and watching them throw. Two committed tests asserted the deleted map's existence and contents. They were pinning the thing being removed, so the tests moved rather than the production code: the 11 role-value pairs survive as a test-local PRE_3897_CODEX_AGENT_SANDBOX baseline, and the assertions now drive the real derivation against real agents/*.md. The coverage is preserved; only its source moved out of production code. validate agents gains checkCodexSandboxPosture, mirroring the existing checkCodexModelPosture: each installed TOML's sandbox_mode must equal the role's expected value, failing with role, expected and found. It previously checked file presence and manifest completeness only, so a TOML whose sandbox_mode disagreed passed. Rung 4 — shortFormToId, recovered rather than invented. I nearly reported this as another wrong §8.3 claim: git log -S returns only documentation commits. Wrong instrument. sdk/src/query/phase.ts at 11918dcc3^ carries five occurrences, and the implementation here matches that code rather than a guess at its semantics - including first-write-wins on a duplicate short form, deterministic from the sorted plan order. It resolves the bare plan number: depends_on: ["01"] now reaches 26-01-auth-hardening. That is a control-flow change, not a diagnostic one - plans that silently collapsed into a single wave 1 now execute in their declared waves, and execute-phase.md consumes those wave values. In-phase only, by construction: the map is built from this phase's rawPlans, so a same-named short form in another phase does not resolve. #3785's display-mapping passthrough and #3885's unresolvable-token warning and wave-verdict suppression are untouched and stay green. If the third tier had over-reached, those are what would have caught it. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): close a fail-open I introduced, and wire the posture check to its command Two blockers from review. Both are mine, and one is a security regression my own change created. 1. A held role could escape its hold by editing its own frontmatter. The Codex install loop set the sandbox identity from the agent's frontmatter `name:` field rather than from its filename, so the hold lookup keyed off a value the file itself declares: deriveCodexSandboxMode('gsd-doc-writer', <real file>) -> read-only deriveCodexSandboxMode('gsd-doc-writer-x', <same file, name: edited>) -> workspace-write deriveCodexSandboxMode('GSD-Doc-Writer', <same file, name: recased>) -> workspace-write What makes this a blocker rather than a nit is the DIRECTION. The deleted CODEX_AGENT_SANDBOX map had the identical lookup-key quirk, but it was an allowlist: an unmatched key fell back to read-only, which is safe. The new scheme derives workspace-write from the tool contract and uses the hold as a subtraction, so the same mismatch fails OPEN. I converted a fail-closed quirk into a fail-open one and did not notice; the isolated reviewer proved it by execution. Neither safety net caught it. validateCodexSandboxHolds only checks that <key>.md exists, never that a file's derived identity matches its key. checkCodexSandboxPosture looks the canonical source up by the installed TOML's filename, finds nothing for a renamed agent, and treats it as a custom non-roster agent — silently no violation. The identity is now the FILENAME STEM, which is what validateCodexSandboxHolds already validates and what an attacker editing frontmatter cannot change without renaming the file — at which point the existing validator catches it. The lookup is case-insensitive so a recase does not slip past either. The frontmatter name still drives the TOML body and filename, unchanged; only the sandbox identity moved. All 35 roster files were checked: name matches filename stem everywhere, so a stricter "they must agree or throw" invariant would have been safe against real content. It is deliberately NOT added — it would abort an install on a tampered file where emitting a correctly-derived read-only TOML is the safer outcome. Recorded as a fork rather than decided silently. 2. checkCodexSandboxPosture was exported and never called. cmdValidateAgents (src/verify.cts) called checkAgentsInstalled and checkCodexModelPosture only; grep for the sandbox check in that file returned nothing. So criterion 3 — "validate agents fails on semantic drift, not only on missing files" — was unmet, and `validate agents` behaved exactly as before. That is ADR-3473 Decision 2's named shape: a declared policy with no executor. It also meant the T28 test asserted at the helper's return value while the COMMAND stayed broken — the ADR-3180 Decision 4(b) failure this epic exists to close, committed by me while enforcing it elsewhere in the same epic. Now wired as an additive `sandbox_posture` field beside `codex_posture`, following the sibling precedent exactly. Drift is report-only, not a non-zero exit, because that is what checkCodexModelPosture does — two sibling posture checks disagreeing about whether a violation is fatal would be its own defect. The choice is recorded in a comment rather than left implicit. A consumer-output test now drives the real CLI and asserts on the emitted JSON, and was shown failing before the wiring and passing after. Also corrected a stale artifact: the design's Known limit L1 still claimed rung 3 was not in this deliverable, written while it was halted and false once the maintainer unblocked it. Verified after both fixes: the three bypass probes all return read-only, the per-role table is 35/35 byte-identical, and both hold self-invalidation cases still throw. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): the marker rung, the derived sandbox, and the bare plan-number depends_on Reference: the runtime precedence ladder in docs/CLI-TOOLS.md gains the install marker rung; docs/COMMANDS.md documents validate agents' new sandbox_posture field; docs/reference/plan-md.md documents that depends_on accepts the bare plan number. Explanation: a docs/features fragment keyed id 3897, so it cannot collide with a concurrent PR hand-allocating a section number, regenerated into FEATURES.md. ADR-3473 §8.3 gains an ANSWER blockquote in the document's own correction style, recording what was measured and built against the section's 2026-08-26 correction - including the qualification that checkAgentsInstalled itself still checks presence only, and the semantic assertion lives in a sibling wired into validate agents rather than folded into it. No how-to. Both user-visible changes are zero-step: a non-Claude install resolving its own runtime, and plans executing in their declared waves, both happen without the user doing anything. docs/how-to/control-the-reported-host-runtime.md covers a DIFFERENT ladder (resolveReportedRuntime / agent_runtime) that this change does not touch, and was deliberately left alone rather than edited by association. No tutorial - nothing multi-step to walk through. docs/AGENTS.md unchanged: it documents Claude-side tools frontmatter, never Codex sandbox_mode, and the emitted tools contract did not change. The prompt layer documents depends_on only by example, not by schema, so nothing there needed editing - and few-shot-examples/plan-checker.md already showed depends_on: ['01'], which now actually resolves. Translated copies of plan-md.md are untouched; the project treats translations as community-maintained. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): move the sandbox derivation out of the installer, off the install path, and off a third parser The full suite came back with 26 failures across four files. Three distinct causes, mapped individually rather than assuming the first explained the rest. A. Requiring bin/install.js printed the GSD banner to stdout and corrupted `validate agents` JSON. Unexpected token '', "[36m ██"... is not valid JSON checkCodexSandboxPosture reached deriveCodexSandboxMode by lazily requiring bin/install.js, whose module load prints the ASCII banner. So the command emitted banner bytes before its JSON and every JSON consumer broke, including ten tests that predate this branch. src/ reaching into bin/ was backwards layering that happened to also be loud. The derivation now lives in src/codex-agent-toml.cts - the existing Codex TOML domain module, no new module and no six-gate ripple - and both bin/install.js and src/agent-install-check.cts import it. One owner, which is §8.3's rule applied to the fix for §8.3. B. The stale-hold throw fired on a legitimate partial source dir, and masked a security assertion. validateCodexSandboxHolds treated "this hold's .md is absent from the install SOURCE dir" as a stale hold and threw. A test fixture, or any partial install source, legitimately contains a couple of agents. Worse, it threw BEFORE the path-escape check, so a test asserting that a `../../evil` frontmatter name is rejected got my unrelated error instead of the traversal rejection it was written for. A fail-closed check of mine was hiding a real security check. The "no stale holds, shrink-only" invariant is a property of the repo's canonical agents/ roster, not of whatever directory an install happens to read. It is off the runtime path and enforced where it belongs, in the tests that already existed for it. A partial source dir now installs cleanly, and the evil-name case throws with its own escapes-configHome message again. C. T8 depended on ambient process.env state. The marker/env parity assertion round-tripped through live process.env. It now compares against resolveExplicitRuntime's already-exported dependency-injection parameter - deterministic and hermetic, same claim. Proven still falsifiable rather than assumed: with the marker rung's normalization temporarily bypassed the two rungs diverge ("codex\n../../etc/passwd" vs "codex-../../etc/passwd") and the assertion fails, then passes again once reverted. One correction folded in along the way. The first version of the move added private _extractFrontmatterAndBody/_extractFrontmatterField helpers to codex-agent-toml.cts - a THIRD copy of frontmatter extraction, where the graph already shows two (bin/install.js:2348, runtime-artifact-conversion.cts:893). Adding a third inside the epic whose thesis is one implementation per rule is not defensible. deriveCodexSandboxMode no longer parses anything: it takes (identity, toolsValue) and each caller supplies the tools value using the extractor it already has. Both helpers are deleted. The identity argument is still the filename stem, so the fail-open fix is untouched. Verified after all three: `validate agents --raw` emits parseable JSON with no banner and both posture fields; the four hold-bypass probes still return read-only; the per-role table is 35/35 byte-identical at 26 read-only / 9 workspace-write; the hold list is still 16. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): drop a dev-only transitive dep, make the derivation total, retire a stale fallback test Suite down to 7 failures from 26. Three more causes, mapped individually. A. My extractor import dragged in a script that does not exist in an installed tree. Cannot find module '../../../scripts/fix-slash-commands.cjs' Chain: src/agent-install-check.cts imported runtime-artifact-conversion.cjs, which requires command-roster.cjs, whose line 36 requires ../../../scripts/fix-slash-commands.cjs. That path exists in the repo and not in an install, so every test exercising a synthetic install dir died at module load. I picked that extractor for convenience without checking what it pulls in - the same mistake that produced the banner bug, one layer further out. agent-install-check now uses a single-purpose extractToolsLine on codex-agent-toml.cts. That is deliberately NOT a general frontmatter parser: we deleted those helpers a commit ago for good reason, and this reads one line. Verified from outside the repo root that requiring either module prints nothing and does not throw. B. A test pinned the deleted name-based fallback. 'defaults unknown agents to read-only' called generateCodexAgentToml with a fixture declaring tools: Read, Write, Edit. Under derivation an unknown agent with a writing contract correctly derives workspace-write - design row S6, a new writing role gets the contract, not the pin. The behavior it asserted was the silent fallback this rung deleted; identity no longer decides the sandbox. Replaced with two rows rather than a flipped string: no tools declared -> read-only (absence is not a grant), and Write/Edit declared -> workspace-write. Strictly more coverage than the row it replaces. C. The stale-hold check still threw per derivation call. Last commit took the roster-existence check off the install path, but deriveCodexSandboxMode itself still threw when a hold's role did not derive broader FOR THE CONTENT IT WAS HANDED - so it fired on any synthetic fixture for a held role. The throw is gone, and it cost nothing: if a held role's content does not derive broader, the hold pins read-only and derivation returns read-only anyway, so the hold is a no-op and there is nothing to fail about. The staleness invariant is a property of the real agents/ roster, and validateCodexSandboxHolds still enforces it there - confirmed against the real roster after the change, not assumed. deriveCodexSandboxMode is now total: every (identity, toolsValue) including undefined and null returns read-only or workspace-write, never throws. Verified: validate agents emits parseable JSON; the four hold-bypass probes return read-only; the per-role table is 35/35 at 26 read-only / 9 workspace-write. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): put the rung-3 decision in the shipped docs instead of pointing at an ignored path The ADR entry and the feature fragment both ended their rung-3 explanation with "see .gsd/phase/feat-3897-adr3473-83-rungs/45-decision-rung3-sandbox.md". That directory is gitignored (.gitignore:55), so the rationale for holding 16 roles at read-only was reachable only from the machine that produced it. A reader of the ADR got a pointer to nothing. Both now carry the reasoning inline: the criterion asks both that the sandbox derive from the declared tool contract and that no role gain a broader sandbox, and those cannot both hold, because a faithful derivation widens 16 roles the deleted map never listed and that fell through its silent read-only default. The resolution is derive-and-hold - the derivation owns the rule now, each hold is released as its enforcement question is answered, and a hold is reversible where a widened sandbox that turns out to be enforced is not. Checked before assuming this was a defect class: CONTEXT.md cites .gsd/phase/<slug>/40-design.md as its standard Design: provenance line in eight module entries, and four other shipped docs do the same. Citing a phase artifact is an established convention here, so those are left alone. What was wrong was specific to these two: they put load-bearing rationale behind the pointer instead of provenance. docs/FEATURES.md regenerated from the fragment via scripts/gen-features.cjs rather than hand-edited. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): close a fail-open, stop a silent mis-resolution, and read a declaration as a declaration Two orthogonal reviews on the shipped sha. Three of the findings are the same failure class this epic exists to close, committed inside it. 1. BLOCKER - the sandbox was decided for one identity and applied to another. bin/install.js derived sandbox_mode for the filename stem and then wrote the result to `${name}.toml`, where name comes from the file's own frontmatter. Make the two disagree and a HELD role's artifact goes wide: rename gsd-doc-writer.md -> gsd-doc-writer-v2.md, keep name: gsd-doc-writer -> stem is unheld, derives workspace-write, lands on gsd-doc-writer.toml add any gsd-*.md whose frontmatter name: is a held role -> clobbers that role's toml with workspace-write Both emit read-only on origin/next, because the deleted map was an allowlist and a miss fell back safe. This is a regression my change introduced. The previous review round moved the HOLD KEY off frontmatter to the filename stem and left the OUTPUT PATH on frontmatter; my own comment at install.js:6985 calls that value attacker-editable, four lines above the line that uses it as the filename. The decision is now made over BOTH candidate identities, most-restrictive wins: if either the stem or the emitted name is held, the mode is read-only. 2. MAJOR - hold matching was toLowerCase() only, so confusables escaped. Turkish dotted/dotless i, fullwidth, NFD, trailing space/NBSP/dot/newline, ./ and ../agents/ all slipped the hold and emitted workspace-write. Identities are now basenamed, trimmed of NBSP/zero-width/control characters, NFKC-normalized and lowercased - and anything still carrying a character outside [a-z0-9._-] is treated as suspicious and derives read-only. We do not enumerate confusables; every shipped roster file is ASCII, so refusing to widen on an identity we cannot recognize is fail-closed with no false positives on real content. 3. MAJOR - the short-form depends_on tier mis-resolved SILENTLY. shortFormToId keyed on the last dash-segment of any canonical id with no constraint that it is a plan number, so a phase holding 09-FIX-auth-PLAN.md made depends_on: ["auth"] bind at wave 2 with zero warnings. This is the worst shape in the epic: the unresolvable-token warning fires on a DROPPED token, so a MIS-RESOLVED one is invisible and the tool reports a confident wave assignment built from a wrong edge. A wrong edge is worse than a missing one. The segment must now match /^\d+$/, which is exactly the contract docs/reference/plan-md.md already documents. This tier was recovered verbatim from the retired SDK lineage, which carried the same defect; we are deliberately NOT preserving it bug-for-bug, and the comment says so, so the next reader does not "restore" it. 4. MAJOR - the derivation was reading a declaration as an absence. extractToolsLine read one line, so a YAML list-form tools: block returned only its first item. Two roster files use list form, and gsd-nyquist-auditor declares Write and Edit there - parsed as "- Read", found no write tool, and emitted read-only. Rung 3's headline claim is that sandbox_mode derives from the declared tool contract; that claim was false for 2 of 35 roles and materially wrong for 1. Reading a declaration as an absence is the silent-drop class this epic exists to close. Renamed extractToolsValue and taught it both shapes. gsd-nyquist-auditor now derives workspace-write and joins CODEX_SANDBOX_HOLDS as its 17th entry, per the standing derive-and-hold decision - so emitted TOML stays byte-identical at 26 read-only / 9 workspace-write while the hold list finally records every role that would widen. A previous pass declined this fix because it moved the count; that inverts the priority. Byte-identity is preserved THROUGH the hold, not by leaving a parser broken. Divergence check, because this is where that bug hides: both paths feeding sandbox derivation - install.js's emitter and checkCodexSandboxPosture - now route through the one extractor. The tools readers in runtime-artifact-conversion and install.js's other frontmatter call sites serve Claude-side emission and do not feed sandbox derivation. Also fixed, each real: the posture check's `found` used a naive whole-file regex where its own sibling uses the block-aware scanner, so prose inside developer_instructions produced a false violation; `found` skipped truncatePostureValue and leaked a 300-char value into validate agents output; deriveCodexSandboxMode's absolute never-throws claim was false for an object with a throwing toString; T49 could not falsify cross-phase leakage (its target phase had its own 01, so a globally-scoped map passed too); T20/N6 iterated a hardcoded table and pinned the FIXTURE size, so a 36th agent would be silently unchecked; three tests reimplemented the code they were testing instead of importing it; and T2-T4 deleted GSD_RUNTIME without restoring it. Verified: hold list 17, gsd-nyquist-auditor derives workspace-write unheld and emits read-only held, roster 35/35 at 26/9, depends_on ["auth"] no longer resolves while ["01"] still does, both identity-bypass cases and every confusable vector emit read-only. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): the hold list is 17, and the reason the 17th was missing The count read 16 because the derivation could not read the declaration it claimed to derive from: the tools reader was single-line, so a YAML list-form tools: block returned only its first item and gsd-nyquist-auditor's declared Write and Edit were read as an absence. Both the ADR entry and the feature fragment now carry the corrected count and the reason for it, rather than a silently updated number. Deriving from a declaration you cannot parse is not deriving, and a flattering count is worse than a wrong one because it looks settled. docs/FEATURES.md regenerated from the fragment. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3897): backfill changeset pr number Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Architecture Decision Records
This directory contains Architecture Decision Records (ADRs) for GSD.
Each ADR documents one architectural decision: what was decided, why, and what consequences follow. ADRs are append-only. Amendments extend existing ADRs with a dated section rather than replacing them.
Reading this corpus
Start with the index below, and respect the status. The index is grouped so that the first table — Active decisions — is the set that governs the system as it stands. An ADR in Superseded, Retired, and Legacy is historical: it records what was once decided and names what replaced it. Do not cite it as current architecture.
Two things the index makes explicit, because getting them wrong has actually misled readers here:
- "Read first" on an active ADR points at a broader ADR that now frames it. A decision can be entirely correct and still not be the whole picture. The runtime capability descriptor (ADR-1016) is live and load-bearing, but ADR-1239 (EoS — GSD as an Embeddable Orchestration Engine) subsumes it as the declarative adapter and inverts its direction: GSD is the engine a host embeds, not an installer that projects onto a host. For how GSD meets a host, EoS is the current frame.
Proposedmeans not ratified — and it is kept honest. On 2026-07-17 the corpus was audited against the shipped tree and nine ADRs whose decisions had demonstrably shipped were ratified toAccepted, each carrying a dated Ratification section with the evidence (see ADR-857 for the fullest example). The ADRs that remainProposedareProposedfor a reason recorded in the file — an unmet acceptance criterion, an outstanding phase, or a successor ADR already planned — not through neglect. Trust the label; if you think it is wrong, prove it in a dated section and see Ratifying a staleProposed.
Naming Convention
New ADRs use issue#-prefix slug naming:
docs/adr/<issue#>-<kebab-slug>.md
Examples: 2264-golden-parity-redesign.md, 1239-gsd-embeddable-orchestration-engine.md.
Why
Two developers computing "next ADR number" locally against main will independently pick the same integer and both ship. The collision is already on disk — 0010-* exists twice and 0011-* exists three times. GitHub issue numbers are server-assigned and atomic: the moment you open an issue, that number is reserved globally. Two PRs that both edit the ### Fixed block of CHANGELOG.md always conflict on merge — two PRs that each use a distinct issue# as their ADR prefix never collide. Same shape, same solution.
Legacy naming is not Legacy status
Files 0001-* through 0012-* are preserved as immutable historical record of the old local-compute numbering. The duplicate 0010-* and the three-way 0011-* are documented residue of that convention — not patterns to imitate. Do not renumber them.
This is the single authoritative statement of the legacy range.
docs/contributor-standards.mdreferences it rather than restating it, so the two cannot drift.
Two other zero-padded files look legacy but are not: 0174-retire-gsd-sdk-package-boundary.md (issue #174) and 0656-research-module-seam.md (issue #656) are mis-padded modern ADRs — modern, issue-numbered files whose four-digit padding is a mistake. They are NOT part of the legacy sequential set above and are not "old local-compute numbering" residue.
This is a statement about filenames only. Many of those ADRs are Accepted and load-bearing today (ADR-0002, ADR-0004, ADR-0008, ADR-0009). An old filename says nothing about whether a decision still holds. The Legacy status in the table below is a separate claim — see the vocabulary.
Because 0010-* and 0011-* each resolve to more than one file, a bare cross-reference like "ADR-0011" is genuinely ambiguous. Link the file (see Lifecycle rules).
Full process
See CONTRIBUTING.md — "Proposing an ADR or PRD" for the end-to-end workflow: opening the issue, waiting for approval, naming the file, and submitting the PR.
PRDs live in docs/prd/, not here. (0011-review-default-reviewers-prd.md predates that directory and is kept in place as frozen historical record.)
Lifecycle rules
These are enforced by scripts/gen-adr-index.cjs, which runs in CI via npm run lint:generated-sync. A violation fails the build with the exact file and fix.
1. Every ADR declares one status from the canonical vocabulary
The first word of the Status field must be one of:
| Status | Means | Obligation |
|---|---|---|
Accepted |
Decided and in force. Cite it. | — |
Proposed |
Decided in principle, not ratified. Do not cite as settled. | If the work has demonstrably shipped, ratify it (below) — do not leave the label lying. |
Superseded |
A specific newer ADR replaced this decision. | Must name the successor as a file link. |
Retired |
What this ADR decided no longer exists at all, and no single ADR replaced it. | Say what was removed and when. |
Legacy |
Frozen historical record, kept for provenance; not a pattern to follow. | Say why it is frozen. |
Prose may follow the token (Superseded by [ADR-0174](0174-retire-gsd-sdk-package-boundary.md) (2026-05-23); originally Accepted (2026-05-09)). Both the bullet form (- **Status:** Accepted) and the table form (| **Status** | Accepted |) are accepted.
2. Cross-references to other ADRs are file links, never bare ids
Write [ADR-0011](0011-skill-surface-budget-module.md), not ADR-0011. Bare ids are ambiguous for 0010/0011, and unlinked references cannot be checked.
If you mean an issue, write #857 — not ADR-857. (An ADR and its owning issue often share a number; that is intentional and not a conflict.)
3. Supersession and subsumption are symmetric
These are different relations. Do not conflate them:
Supersedes/Superseded by— the target is replaced. Its status becomesSuperseded.Subsumes/Subsumed by— the target still holds, but a broader ADR now frames it. Its status is unchanged; it becomes a component of the larger decision.
If A declares either relation toward B, B must record the reciprocal. A one-way pointer is the failure this corpus actually suffered: ADR-1239 declared it subsumed four ADRs, none of which said so, and none of which pointed back — so a reader landing on any of them concluded the superseded frame was the way forward.
Only an Accepted ADR is owed the back-link. A Proposed ADR's claim is prospective: it has not taken effect, so its target is not marked. On ratification, the check begins demanding the back-links.
4. The declared id matches the filename
An H1 of # ADR-0175: … in a file named 218-*.md is a rename that never finished. The id in the title must match the filename's prefix.
5. A trailing H1 status bracket must agree with the Status field
Many ADRs restate their status in the H1 — # ADR-1610: … [Accepted]. That bracket is the first thing a reader sees, and the index strips it when rendering the title, so a stale one used to be invisible to everyone but the reader it misled.
If the H1 ends in a bracket holding a status token, it must name the same status as the Status field. Comparison is case-insensitive and against the parsed token, so [Superseded] agrees with Status: Superseded by [ADR-0174](0174-retire-gsd-sdk-package-boundary.md) (2026-05-23).
A trailing bracket that is not a status token — [Draft], [WIP] — is treated as part of the title and left alone. If you want a bracket the gate ignores, do not spell it like a status.
6. Every relative link resolves
A link whose target does not exist on disk fails the check, naming the file, the line, and the unresolved target. This covers every markdown file in this directory, including this README and any file whose name breaks the convention above.
| Written as | Treated as |
|---|---|
[t](900-beta.md), [t](../prd/) |
resolved — a directory counts |
[t](900-beta.md#section) |
the file is resolved; the #fragment is not checked |
[t](https://…), [t](mailto:…), [t](//host/x) |
out of scope — absolute destinations are never fetched |
[t](#lifecycle-rules) |
out of scope — a same-document anchor is not a file reference |
[t](/docs/adr/x.md) |
resolved against the repository root, as GitHub does |
a link inside a ``` fence or `backticks` |
not a link — markdown does not render one there, so it is never resolved |
[text][ref] reference-style, <a href>, bare autolinks |
not supported; write an inline link |
Two consequences worth stating outright:
- Case matters, on every platform.
[t](0001-Alpha.md)pointing at0001-alpha.mdfails even on macOS and Windows, because it 404s on github.com and reds the Linux CI lane. The failure names the entry it found so the fix is obvious. - A link to a generated or ignored path fails. Nothing here consults
.gitignore; the question is only whether a reader following the link lands somewhere. Cite the hand-authored source rather than the build artifact.
If the gate rejects something you wrote
Reproduce it locally first — it is the same command CI runs, and it names the file, the line, and the target:
node scripts/gen-adr-index.cjs --check
Then work from the reason:
| What it says | What to do |
|---|---|
does not resolve — no such file or directory at … |
Fix the path. It is relative to docs/adr/, so a sibling ADR is just 900-slug.md. If the target genuinely does not exist yet, drop the link rather than leaving it pointing nowhere. |
…Did you mean X? — link targets are case-sensitive on github.com |
Match the on-disk name exactly. Your machine may open the file regardless; github.com and the Linux CI lane will not. |
escapes the repository |
The path resolves outside the repo. Link something inside it, or use an absolute URL — those are out of scope and never checked. |
is a symlink that escapes the repository |
An ADR file itself is a symlink pointing outside the repo. Commit a real file. |
H1 status bracket […] contradicts the Status field (…) |
Update whichever of the two is stale so they agree. The Status field is authoritative; the bracket is a restatement for the reader. |
A link that is an example, not a destination, belongs in backticks. The gate skips fenced blocks and inline code entirely, because markdown does not render a link there. That is the escape hatch for illustrative syntax — the table above is written that way, which is why it does not fail this check. An indented code block (four spaces) is not skipped; use backticks.
To consume the result from a script rather than by eye, use --json (below) and branch on each violation's stable reason code.
Ratifying a stale Proposed
A stale Proposed is not cosmetic: it tells contributors and agents that live architecture is an unbuilt idea. Fix it — but on evidence, not vibes.
The bar. All four must hold before flipping to Accepted:
- The decided mechanism demonstrably exists in the tree — name the files, symbols, and tests.
- The owning issue is closed as completed. A closed issue is not proof:
stateReasonof not planned / duplicate means the decision was dropped (that isLegacyorRetired, notAccepted). - No material part is unshipped. If the ADR defines phases and one is outstanding, or states its own bar for acceptance and that bar is unmet, it stays
Proposed. - No later ADR supersedes it, and no approved issue already plans its graduation as separate work.
The procedure. Set the status to Accepted — ratified <date> (originally Proposed <date>), add a dated ## Ratification section holding the evidence, then run node scripts/gen-adr-index.cjs --write. If the ADR claims to supersede or subsume others, the gate will now demand their back-links — that is the point. Ratify deliberately.
Two traps worth knowing, both hit during the 2026-07-17 audit:
- Shipped code is necessary, not sufficient. Eight ADRs had every named module, symbol, and test present and their epics closed — and still failed the bar: ADR-2264's own headline acceptance criterion is unmet in the tree, ADR-230's decided branch protection does not match the live API, ADR-660's namesake mechanism is performed by hand, and ADR-959 has an approved issue planning its graduation as its own ADR. Verify the decision, not just the code.
- "Supersedes" is often "subsumes". Read what the ADR means before the gate makes you act on what it says. ADR-857 said "Supersedes (generalizes)"; taken literally, ratifying it would have stamped two live seams (ADR-0011, ADR-58) as dead. The parenthetical was the truth; the field name was wrong.
Maintaining the index
The index is generated. Do not hand-edit it. Everything between the ADR-INDEX:START / ADR-INDEX:END markers is derived from the ADR files themselves:
node scripts/gen-adr-index.cjs # print the index
node scripts/gen-adr-index.cjs --write # regenerate it into this file
node scripts/gen-adr-index.cjs --check # CI: fail if stale or invalid
node scripts/gen-adr-index.cjs --json # same checks, machine-readable report
After adding an ADR, or changing any ADR's status or relations, run --write and commit the result. npm run lint:generated-sync runs --check in CI, so a missing or stale row fails the build rather than rotting silently.
--json runs the same validation as --check and writes a report to stdout instead of prose to stderr, with the same exit code. Each violation carries a stable reason code, so a tool consuming this never has to pattern-match an error message:
{
"ok": false,
"adrCount": 76,
"indexStale": false,
"violations": [
{ "file": "2704-example.md", "line": 41, "reason": "link_unresolved",
"target": "reference/x.md", "resolved": "docs/adr/reference/x.md" }
]
}
An unrecognized flag is rejected rather than ignored.
This replaces a hand-maintained table that had drifted to 40 of 65 ADRs — the entire capability family and EoS itself were missing from it, which is precisely why the ADRs a reader most needed were the ones they could not find.
Index
Active decisions
These govern the system as it stands. Cite these.
| ADR | Title | Status | Read first |
|---|---|---|---|
| ADR-0001 | Dispatch policy module as single seam for query execution outcomes | Accepted | — |
| ADR-0002 | Command Contract Validation Module | Accepted | — |
| ADR-0003 | Model Catalog Module as single source of truth for agent profiles and runtime tier defaults | Accepted | — |
| ADR-0004 | Planning Workspace Module as single seam for worktree and workstream state | Accepted | — |
| ADR-0006 | Planning Path Projection Module for SDK query handlers | Accepted | — |
| ADR-0008 | Installer Migration Module owns install-time upgrade safety | Accepted | — |
| ADR-0009 | Shell Command Projection Module owns runtime-aware OS command rendering | Accepted | — |
| ADR-0011 | review.default_reviewers config key scopes the no-flag /gsd-review fan-out |
Accepted | — |
| ADR-0011 | Skill Surface Budget Module owns install-time profile staging and runtime surface control | Accepted | ADR-857 |
| ADR-15 | Cross-AI Plan Convergence via Existing Orchestration Commands | Accepted | — |
| ADR-22 | Plan-vs-codebase drift guard: defaults and symbol-resolver seam | Accepted | — |
| ADR-58 | Runtime Install Policy Module owns the typed install-plan projection | Accepted | ADR-1239, ADR-857 |
| ADR-0174 | Retire @opengsd/gsd-sdk package boundary — single-runtime collapse | Accepted | — |
| ADR-218 | Harden release-workflow version validation — reject leading zeros and pre-check npm | Accepted | — |
| ADR-227 | Input validation must check semantic shape, not just type | Accepted | — |
| ADR-415 | Prevent stale-base reintroduction of retired runtime tokens | Accepted | — |
| ADR-443 | Unified cross-provider effort controls and fast-mode-aware routing | Accepted | — |
| ADR-452 | Adopt standard ESLint flat-config lint harness | Accepted | — |
| ADR-456 | Test-rigor architecture — deterministic scheduling, antagonistic tier, typed-surface mandate, and delete-bad-tests policy | Accepted | — |
| ADR-457 | Generation model for bin/lib/*.cjs type safety |
Accepted | — |
| ADR-550 | spec-phase probe pattern and prohibition contract | Accepted | — |
| ADR-0656 | Research Module — L2-hybrid seam for cached, curated-first research | Accepted | — |
| ADR-766 | Claude Code Plugin Manifest Module owns the projection of gsd-core surfaces onto the Claude Code plugin contract | Accepted | — |
| ADR-857 | Capability system — five-step loop as core, features as plug-ins behind Loop Extension Points | Accepted | — |
| ADR-894 | Capability declaration format + registry generation | Accepted | ADR-1239 |
| ADR-959 | Capability Command Contribution | Accepted | — |
| ADR-1016 | Runtime Capability Descriptor | Accepted | ADR-1239 |
| ADR-1235 | Migrate agent conversion to the descriptor-driven install path | Accepted | — |
| ADR-1239 | GSD as an Embeddable Orchestration Engine | Accepted | — |
| ADR-1244 | Capability Ecosystem: third-party authoring, versioned manifests, and URL import/upgrade/remove | Accepted | — |
| ADR-1372 | Canonical markdown-structure parsing — the markdown-sectionizer seam |
Accepted | — |
| ADR-1411 | Resolution must report provenance, not fall open silently | Accepted | — |
| ADR-1508 | Runtime Artifact Conversion Module owns per-runtime content rewriting | Accepted | — |
| ADR-1517 | Reviewer instances — bounded config surface for same-adapter multi-model review | Accepted | — |
| ADR-1577 | Untrusted-input boundary + opt-in injection blocking | Accepted | — |
| ADR-1593 | Skill mapping & converter methodology across runtimes | Accepted | — |
| ADR-1610 | workflow & agent size-budget ratchet (per-file byte baseline + tier hard caps) | Accepted | — |
| ADR-1703 | Cross-platform portability enforcement as AST ESLint rules | Accepted | — |
| ADR-1769 | STATE.md Transition Module — intent-based transitions over scattered RMW callbacks | Accepted | — |
| ADR-1787 | /gsd:next smart-entry front door delegates advancement to /gsd:progress --next |
Accepted | — |
| ADR-1817 | STATE.md rebuild — derivability contract (capstone transition) | Accepted | — |
| ADR-1820 | Spec-Optional Predicate Rail — the Spec-Section Detection Module, the fallback toggle, and the SPEC↔probe precedence contract | Accepted | — |
| ADR-1866 | agent_skills dual injection — orchestrator-side + agent-side self-load | Accepted | — |
| ADR-1990 | Existing Code Onboarding Module owns deterministic repo-state detection and onboarding route selection | Accepted | — |
| ADR-2008 | Generic gate-predicate evaluator | Accepted | — |
| ADR-2121 | Phase-Identifier Parsing Consolidation | Accepted | — |
| ADR-2143 | Markdown Table Model, Bounded Mutation, and Fail-Loud Consolidation (#1372 part 2) | Accepted | — |
| ADR-2164 | Statusline draws its data boundary at local, read-only sources | Accepted | — |
| ADR-2207 | STATE.md Status lifecycle — phase-completion writes an intermediate state; milestone-close owns termination |
Accepted | — |
| ADR-2313 | Codex Adopts the Passive / Session-Only Model Posture | Accepted | — |
| ADR-2346 | Command Dispatch Completion | Accepted | — |
| ADR-2363 | A capability's skill body is an instruction surface — trusted, unscanned, and disclosed | Accepted | — |
| ADR-2619 | Observability and shareable diagnostics — wire the dispatch seam, add the outbound trust boundary | Accepted | — |
| ADR-2629 | Phase effort is estimated against a calibrated smart-zone budget, not a static heuristic | Accepted | — |
| ADR-2719 | Emitted-artifact attribution — replace the committed parity fixtures with a computed conservation law | Accepted | — |
| ADR-2782 | Reviewer Lane — the cross-AI reviewer handoff becomes a declared capability surface | Accepted | — |
| ADR-2866 | Install-surface resolution — the install pipeline resolves (runtime × scope × trigger) as a value |
Accepted | — |
| ADR-2966 | Test the five-step loop as a continuous walk, not isolated points | Accepted | — |
| ADR-2980 | A payload-carried error key is a degraded result, not a fault |
Accepted | — |
| ADR-3180 | Planning Semantic Model — Single Owner per Derivation | Accepted | — |
| ADR-3212 | The Lexical Seam — Safe Pattern Construction, Line-Terminator Normalization, and Tokenizer-First Stateful Grammars | Accepted | — |
| ADR-3408 | STATE.md Write Path — One Declared Policy, One Write Seam | Accepted | — |
| ADR-3409 | Shell Guards Must Observe Their Own Failure Arm | Accepted | — |
| ADR-3473 | Enforcement by Construction — One Owner per Invariant | Accepted | — |
| ADR-3574 | Install materialization shares primitives, not one writer | Accepted | — |
| ADR-3625 | The platform seam keeps its own Windows binary resolution rather than adopting a spawn library | Accepted | — |
| ADR-3660 | Runtime Artifact Layout Module owns per-runtime artifact placement | Accepted | ADR-1239 |
Proposed
Decided in principle, not yet ratified. Do not cite as settled architecture.
| ADR | Title | Status | Read first |
|---|---|---|---|
| ADR-230 | Introduce next as a long-lived integration branch |
Proposed | — |
| ADR-612 | Bracket Phase-ID Convention | Proposed | — |
| ADR-660 | Release from the head of next; immutable release tags; @next dist-tag as the RC surface |
Proposed | — |
| ADR-1143 | Claude orchestration capability — Workflow tool (ultracode) as a runtime-gated loop execution backend | Proposed | — |
| ADR-1213 | Capability write side — the Capability State Writer | Proposed | — |
| ADR-1606 | prohibition-enforcement verify-time seam | Proposed | — |
| ADR-1671 | Dynamic context management platform | Proposed | — |
| ADR-1953 | Complexity-triggered refactor — the loop measures the entropy it just added | Proposed | — |
| ADR-3128 | Adaptive runtime evidence for GSD Debug | Proposed | — |
| ADR-3889 | One exit-code registry — 0 and 1 are free, everything else is allocated | Proposed | — |
| ADR-3942 | The emitted-drift acknowledgment is PR-lifetime data — it belongs in a commit trailer, not the working tree | Proposed | — |
Superseded, Retired, and Legacy
Historical record. Do not follow these — each names what replaced it, or why it was retired.
| ADR | Title | Status | Replaced by |
|---|---|---|---|
| ADR-0005 | SDK Architecture seam map for query/runtime surfaces | Superseded | ADR-0174 |
| ADR-0007 | SDK Package Seam Module owns SDK-to-get-shit-done-redux compatibility | Superseded | ADR-0174 |
| ADR-0010 | File Operation Engine Module owns safe runtime/config file mutations | Superseded | ADR-0009 |
| ADR-0010 | Skill Surface Budget Module owns install-time skill listing curation | Superseded | ADR-0011 |
| ADR-0011 | PRD — review.default_reviewers config key for /gsd-review reviewer selection |
Legacy | — |
| ADR-0012 | CommandRoutingHub as single dispatch seam for CJS command families | Superseded | ADR-0174 |
| ADR-2264 | Redesign golden-install-parity — single-source manifest builder + split invariant | Superseded | ADR-2719 |
| ADR-3524 | CJS↔SDK hard seam — one source of truth per Shared Module | Superseded | ADR-0174 |
Generated by scripts/gen-adr-index.cjs — run --write after adding or restatusing an ADR.
Seam map
Orientation for the module-ownership ADRs. This section is prose and hand-maintained; the index above is the authority on status.
How GSD meets a host — start at ADR-1239 (EoS). It is the current frame and subsumes the descriptor/projection ADRs (ADR-1016, ADR-58, ADR-3660, ADR-894) as adapters beneath it.
The SDK seam map is gone. ADR-0005 was once the entry point for SDK module ownership; it is superseded by ADR-0174, which retired the @opengsd/gsd-sdk package boundary entirely. There is no sdk/ tree. Read ADR-0174 for the single-runtime collapse; the seam-Module vocabulary survives under one src/.
ADR-0006 documents how query handlers project planning paths (cwd → effectiveRoot → .planning/<project>/...). Cross-reference the Planning Workspace Module (ADR-0004) for workstream pointer policy.
ADR-0008 documents the Installer Migration Module for safe install-time moves, removals, config rewrites, and user-data preservation.
ADR-0009 documents the Shell Command Projection Module seam for runtime-aware projection of installer-owned command text and projection IR. Its Phases 3–4 absorbed the File Operation Engine Module (ADR-0010).
ADR-0011 documents the Skill Surface Budget Module for install-time skill/agent profile staging (--profile=<name>, .gsd-profile marker, requires: closure) and the Phase 2 runtime /gsd:surface command.
ADR-1411 establishes the Resolution Provenance principle: context resolution (config loading, project-root anchoring, workstream resolution) must report its provenance rather than fall open silently to defaults. It is the resolution-side analog of ADR-227 (input-validation shape).