4 Commits

Author	SHA1	Message	Date
Tom Boucher	7e9ce08e2c	test(#2736): property-test parsePhaseFromProse name precedence (#2878) ... #2821 reworked parsePhaseFromProse's NAME precedence (dash-vs-paren choice, status-keyword tails, `Milestone:` tails, paren-stripped separator search, the lone-ALL-CAPS-tail rule) but shipped it pinned only by hand-picked examples. The two pre-existing property blocks cover phase-token ANCHORING (#2111) and "N of M" phase extraction; neither touches name precedence. This adds nine properties over the canonical parser. P1 and P9 are the delta guards: both fail against the pre-#2821 paren-first parser (proved with a standalone mutation harness), because each requires a genuine em-dash name to win over a co-present parenthetical. P9 additionally exercises the paren-stripped separator search, since the losing parenthetical itself contains an em-dash. P2-P4 are characterization tests for precedence contracts both parser versions satisfy; P5-P8 pin totality and phase-token extraction, which #2821 left alone. The section comment states which is which so a future reader does not mistake the characterization tests for delta guards. The generator's status-word exclusion filter is a test-local mirror of the private, unexported STATUSY_TAIL_RE. A divergence-guard test pins that mirror to observable parser behavior (not source text, which the no-source-grep rule forbids), so an implementation vocabulary change fails loudly instead of silently weakening every property that depends on the filter. Also clears six pre-existing no-unused-vars lint warnings surfaced by the lint run for this change (dead bindings in four unrelated test files, deleted rather than underscore-renamed); removing the never-called openCodeBlock helper cascaded to its now-unused fs/path/reviewPath consts in both fold scopes. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-07-30 15:27:52 -04:00
Tom Boucher	3f6b063fbb	chore(#2799): invoke_reviewers and write_reviews iterate declared lanes (#2861) ... * chore(#2799): resolve reviewer lanes into executable invocation plans Phase 5b of ADR-2782. Adds the resolver and runner that let invoke_reviewers iterate declared lanes instead of hand-authored per-CLI bash. Five additive descriptor amendments, each forced by a lane that ships today: - LaneHandler gains 'opencode' — the lane rebuilds its review from assistant text parts of a --format json stream; a plain stdout copy re-breaks #1936. - modelConfigKey — antigravity's key is review.models.agy, not .antigravity, so resolving by slug silently dropped a configured model. - defaultHost/fallbackModel — Phase 4 federated every *_host with a default of empty string; the real fallback only existed in the bash. - args becomes an argv template with a closed four-placeholder vocabulary. Positional splicing produced 'codex --model M -o F exec --ephemeral', which is not a valid invocation: codex injects in the middle, twice. - kimi-code lane, with the bounded command-capability probe (needle --output-format) that tells Kimi Code from the legacy python kimi-cli. Parity gate re-pointed: the workflow-text families it scanned are the text this phase deletes, so they are replaced by descriptor-to-registry parity plus an anti-parity check that no bespoke leg returns. jq, curl and external timeout/gtimeout all drop out of the review path. Refs #2782 * chore(#2799): add review-lane query surface and widen the manifest vocabulary Adds the gsd-tools 'review-lane' route (plan/invoke/sections) the workflow loops over, projects all twelve lanes into their capability manifests, and widens capability-validator for the amendments. opencode admitted to VALID_LANE_HANDLERS under the second arm of the enum's own admission rule: one lane, justified by a documented upstream defect data cannot express (#1936 — the agent can end its turn with zero output tokens and --format default then drops the assistant text entirely). Two bugs caught by an end-to-end stub run and fixed here: - loadConfigResolved returns a provenance wrapper, not the config; using it directly resolved every key to undefined, which reads as 'nothing configured' and silently dropped every model override. - hasBinary used shell:true with an args array (Node 26 DEP0190). Replaced with a PATH scan that spawns nothing at all. Refs #2782 * chore(#2799): iterate declared lanes in invoke_reviewers and write_reviews Replaces the eleven hand-authored per-CLI bash legs with a loop over resolved lanes, and renders REVIEWS.md sections from each lane's declared reviewsSection instead of thirteen hardcoded headings. review.md drops from 1104 lines to 507 (61KB to 28.7KB). Parity gate re-pointed, as agreed: the leg-marker and section-heading families scanned exactly the text this phase deletes, so they are replaced by descriptor-to-registry parity in both directions, plus an anti-parity check that fires if a bespoke leg is ever re-added. Enum, emitting sites and the Object.keys lock moved together. The budget-trim helper is hoisted out of the Ollama leg: it was always lane-agnostic, and any lane may now declare a promptBudgetKey. Refs #2782 * feat(#2799): bind the consented egress host and re-verify it at invocation Completes ADR-2782 D5. Rule 1 was recorded in the ADR as delivered by Phase 3 but was not implemented: ConsentRecord had no host field and nothing in the tree bound one, so this phase's rule-4 comparison had no baseline. ConsentRecord gains an OPTIONAL reviewerHost. Optional is the whole design: isValidConsentRecord does not require it, so every record already on disk stays valid and no re-consent storm fires (D4 rule 5). It is deliberately excluded from disclosureSignature — the loader has no config resolver, so folding a config-derived value in would make loader and lifecycle compute different signatures for the same manifest and re-prompt forever. Install resolves hostConfigKey (falling back to the lane's declared defaultHost, which is what the invocation path uses) and records it. Invocation re-resolves and blocks on mismatch rather than silently redirecting. Absence allows: no record, or a record predating the field, means nothing to compare — denying there would break every existing local-model user on upgrade. Refs #2782 * test(#2799): cover the resolver, runner and handlers; retarget the parity suites Adds the golden invocation-plan table (one row per shipped lane, derived from the bash legs rather than the descriptor types) plus runner coverage for the probe, empty-output policy, the three handlers and the egress check. Retargets the existing suites onto the new contract: descriptor-to-registry parity, the anti-parity check, the opencode handler, and the twelfth lane. Two corrections found by running them: - modelConfigKey was required; that breaks D4 rule 2, since a reviewer manifest authored before this phase would fail validation on upgrade. It is optional, read as null when absent. - the antigravity non-zero-exit test pre-seeded the transcript, which asserted that a STALE entry leaks through — the exact bug the watermark prevents. The spawn now appends, as the real tool does. Refs #2782 * fix(#2799): restore agy --add-dir and the self-report prompt in the handler Retargeting the three legacy reviewer suites off the deleted bash surfaced two real regressions in the port, both #2176: - --add-dir was dropped. Without it agy's permission context never receives the cwd repo, so the agent anchors on its own scratch dir and reviews the plan text in isolation — the exact failure the Review Instructions forbid. It is capability-probed, because an older agy rejects the unknown flag outright and a lane that fails to start is worse than one running on the prompt anchor. - the prompt lost the clause mandating a REVIEWED-WITHOUT-REPO-ACCESS self-report, which is what makes a blind review distinguishable from a grounded one. antigravity now builds its own prompt variant. Also ports the #2073 mode-2 cli.log diagnostic, which was dropped: a pinned model that 404s exits 0 with empty stdout AND an empty transcript, so agy's own log is the only evidence that anything failed. The three suites now assert against the plan and the handler instead of matching fence text, so they no longer need allow-test-rule exemptions. Refs #2782 * docs(#2799): document the declared lanes, the new flag, and dropped prerequisites COMMANDS.md gains --kimi-code and replaces the jq-prerequisite paragraph, which is now false: no lane requires jq, curl or an external timeout. Adds the changed-egress-destination behavior, since a blocked lane is something a user can hit. CONFIGURATION.md records that the model config key is declared per lane rather than derived from the flag — antigravity's is review.models.agy — and adds review.models.kimi-code. reviewer-instances.md now routes an instance through its lane's single invocation seam instead of a copied per-adapter bash block, which is what lets a cross-cutting fix reach instances for free. That required implementing the --model/--agent/--as flags it documents; --model re-resolves through the lane's argv template rather than splicing, so the flag lands where the lane declares it rather than ahead of a subcommand. CONTEXT.md glossary gains both new modules. Refs #2782 * chore(#2799): drop the stale emitted-drift acknowledgment The only entry was #2797's, acknowledging COMMENT-ONLY GROWTH in review.md. That file now shrinks by ~32KB and every emitted hash that moved is attributable to this diff, so the ack no longer explains anything. Removing the last entry means removing the file: its presence is the alarm, and an empty one signals nothing. Verified by deleting it and re-running the attribution and provenance gates plus lint:ci — all green without it. Refs #2782 * docs(#2799): record the Phase 5b vocabulary widenings in ADR-2782 Five additive amendments, each forced by a lane that ships today, plus two corrections the phase had to make rather than work around: - D5 rule 1 was recorded as delivered by Phase 3 and was not implemented, so this phase's rule-4 comparison had no baseline. Recorded because an ADR asserting a rule was delivered is exactly what stops a later phase checking. - The DEFECT.GENERATIVE-FIX gate is re-pointed: its workflow-text families scanned the text this phase deletes. Also records that D7's 'skip the probe where no bounding mechanism exists' carve-out is obsolete — in practice it meant the Antigravity lane ran unbounded on every stock macOS host, which ships neither timeout nor gtimeout. Refs #2782 * fix(#2799): close four defects found by adversarial review Two confirmed bugs, both reproduced before fixing: - resolveLanePlan was not total. An openai-http lane with a missing or non-object invoke dereferenced inv.hostConfigKey and threw, contradicting the module's own documented contract; the spawn branch guarded correctly and the http branch did not. The CLI seam resolves every selected lane in one map, so one malformed overlay manifest would have aborted the whole review rather than dropping its own lane. Guarded, plus a per-lane try/catch at the seam so a throw can never take down siblings. - A reviewer-instance model was silently dropped for any lane declaring modelConfigKey null (cursor, qwen, coderabbit). reviewer_instances validates that cli is a known slug but never that the slug accepts a model, so a user could configure one, get a clean run, and never learn a different model reviewed their plan. Now warns explicitly. Two hardening fixes: - The slug is concatenated into artifact paths, so LANE_SLUG_RE is enforced in the resolver rather than inherited from a validator that does not run on this path — the module documents itself as the overlay-manifest trust boundary, so it should not depend on someone else having checked. - normalizeHost mangled a scheme-less value: new URL('localhost:11434') parses with an empty hostname, so it became 'localhost://11434' and was compared and requested as if real. An empty hostname now means not-a-URL. Also documents the one gap that cannot be closed here: the antigravity watermark is keyed by workspace, so two concurrent reviews of the same repo share a transcript. agy exposes no per-invocation id to filter on, so the handler now states which half of its never-stale guarantee actually holds. Refs #2782 * test(#2799): retarget the remaining eight review.md-asserting suites The remote runner found 37 failures the local sweep missed (it hit the shell's two-minute cap before reaching these). All eight extract per-CLI bash from review.md that this phase deletes; each protects a real invariant, so each is retargeted onto the plan, the runner or the handler rather than removed. Three real defects surfaced by doing so: - effort args never reached ANY lane. model-resolver.cjs exports no resolveExecution, so effortFor silently returned [] every time. Restored by calling the same bounded resolve-execution query the bash legs used — and NOT with --raw, which prints the resolved effort rather than the picked field, so claude got 'low' instead of '--effort low'. - the timeout guidance lost 'a silent empty output is a timeout kill, not a crash' — the operator note that exists because of the Codex 0xc0000142 misdiagnosis. Restored. - the opencode handler dropped EMPTY assistant text parts. The shipped jq was , and only substitutes for false/null — an empty string is truthy in jq and contributed a blank line. Found by a property test shrinking to ['', '']. The opencode property suite no longer spawns jq at all, which deletes the #2099 hang mechanism it was architected around rather than mitigating it. Refs #2782 * fix(#2799): register the two new generated modules, and untrack them The remote runner caught build output committed to git. Both new modules compile from src/*.cts into gsd-core/bin/lib/*.cjs, and every sibling generated that way is gitignored and eslint-ignored (ADR-457) - including Phase 1's own review-lane-descriptor.cjs. Mine were neither, so repo-invariants' "each bin/lib/*.cjs is linted xor ignored according to migration state" failed. Registered both in .gitignore and eslint.config.mjs alongside the Phase 1 module, and dropped them from the index. Nothing about the shipped behaviour changes; the artifacts are rebuilt by build:lib. This is the new-.cts-module registration ripple, and it is the one part of it I had not completed - the CONTEXT.md glossary and the inventory manifest were already done. Refs #2782 * chore(#2799): backfill changeset pr number to 2861 * chore(#2799): backfill changeset pr number to 2861 --------- Co-authored-by: Test <test@example.com>	2026-07-30 12:48:06 -04:00
Tom Boucher	d1e9491fef	feat(#2099): drive GitHub Copilot through the EoS descriptor + multi-event hook bus (ADR-1239) ... Fold Copilot's residual runtime-literal branches onto descriptor-driven hostBehaviors. Several issue premises were inaccurate (verified via research) and deliberately NOT followed: writesSharedSettings/"legacy exclusion list" (per-runtime descriptor data, copilot's false is correct); RUNTIME_CONTENT_DISPATCH.copilot + installSurface==='copilot-instructions' (already descriptor-driven); extendedHookEvents (closed Claude/Gemini enum — hooks extended in code instead); reapply ternary (already folded, kimi #2095). Real folds (all byte-parity — golden byte-identical for every runtime): - src/install-engine.cts + src/surface.cts: the two `.agent.md` filename cutovers (_copyStaged + _syncGsdDir) unified onto hostBehaviors.agentFileExtension via a new exported agentFileExtensionFor() accessor (kills the two-mechanism divergence). - src/runtime-artifact-conversion.cts: applyAgentPathRewrites' copilot skip → hostBehaviors.noPathRewrite:true (antigravity #2096 precedent). - bin/install.js uninstall: the two isCopilot cleanup branches → installSurface=== 'copilot-instructions' gate (symmetric with install-time). - bin/install.js: `!isCopilot` in the two skipSharedHooksInstall checks → hostBehaviors.skipSharedHooksInstall:true (copilot has no shared gsd-*.js hooks). - bin/install.js: three dead legacy inline-agent-loop isCopilot refs removed (copilot ∈ _DESCRIPTOR_AGENTS_RUNTIMES → unreachable; byte-parity proven by clean golden + real reachable-runtime install diffs). isCopilot dropped from 4 destructures. Zero live `runtime==='copilot'`/`isCopilot` branches remain in bin/install.js, install-engine.cts, surface.cts, or runtime-artifact-conversion.cts (AC2 guard scans all four). UPGRADE 1 (multi-event hook bus): buildCopilotHookConfig() now emits preToolUse/ postToolUse/userPromptSubmitted/sessionEnd advisory handlers alongside sessionStart (static inline bash/powershell — deterministic, golden-trackable). Only copilot.json's gsd-session.json hash changes. UPGRADE 2 (background dispatch): surfaced via the negotiated contract only — dispatch.background:true exceeds the declarative-cli baseline and survives negotiation with no downgrade warning. NO .agent.md frontmatter field (copilot has none). MCP companion out of scope (AC4 names only 2 upgrades). Tests: declarative-reference-copilot (adapter/axes/fail-closed + AC2 4-file source-grep guard) + copilot-upgrades (live 5-event hook wiring; dispatch.background negotiation). Matrix EoS note + how-to; changeset (Changed). capability-registry regenerated. Incidental flaky-test RE-ARCHITECTURE (no-defer, maintainer-directed): tests/opencode-review-reconstruction.property.test.cjs spawned ~600 synchronous execFileSync('jq') subprocesses (numRuns:200 × 3 fast-check properties, one jq per generated stream); a single jq freezing on a contended macos-22 CI runner hung the whole unit-test chunk to its 600s kill (this PR's CI). --test-force-exit can't interrupt a synchronous execFileSync, so the cure is to stop spawning per case, not just time-bound it. Re-architected to run the SHIPPED jq program over the whole fast-check corpus in ONE jq process: each generated stream is one compact-JSON array per line in a temp file, `jq -c <PROGRAM>` (no -s) applies PROGRAM to each array (`.` == the array, exactly what production's `jq -rs <file>` sees after slurping) and emits one result per line — empirically byte-identical to the per-stream form across embedded-newline/empty/quote/ unicode/null-drop cases, and file-input (like production) so there's no stdin pipe to deadlock on large I/O. ~600 spawns → 6; coverage unchanged (200-case corpus per property, deterministic seeds) plus explicit boundary/diagnostic example batches. Still property- tests the real shipped jq (no JS reimplementation). Per-call jq timeout retained as a belt-and-suspenders bound. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	2026-07-11 13:10:31 -04:00
Behruz Nassre Esfahani	23254ca5a7	fix(#1936): reconstruct OpenCode review from JSON events (#1992) ... * fix(#1936): reconstruct OpenCode review from JSON events; diagnosable empty-output stub On a large review prompt, OpenCode's default `build` agent runs a few read tool calls then ends its turn with zero output tokens (reason:"stop", output:0), so `opencode run --format default` emits empty stdout. The reviewer block redirected stderr to /dev/null and wrote a generic "failed or returned empty output" stub — so the phase silently lost its second independent reviewer with no diagnostic and no timeout. Rewrite the OpenCode reviewer block to invoke `--format json` as the primary call and reconstruct the review from the assistant `text` parts (jq). Capture stderr to a `.err` sidecar (mirrors the Codex block). When the agent emits no text, surface the stop reason, output-token count, and stderr so the failure is diagnosable. Gate the stub on the extracted CONTENT, not the output file size — an empty jq extraction still prints a lone newline that a `[ -s file ]` check would treat as populated. Document the wall-clock timeout as a Bash-tool param (macOS lacks GNU timeout; opencode has no native timeout flag). review.md was already at the DEFAULT size-tier ceiling (40956/40960), so the fix cannot fit without reclassifying it into the LARGE tier (it is a multi-reviewer orchestration file that outgrew "focused single-purpose"; 43.4 KB sits well under the LARGE high-water mark). Recapture the 16 golden-install fixtures — the diff is exactly one review.md hash per runtime. Regression block folded into review-default-reviewers-workflow.test.cjs (new bug-NNNN test files are not accepted). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(#1936): add changeset * test(#1936): property-test the OpenCode review jq reconstruction Address the re-review's one actionable finding: the jq JSON-event → text reconstruction had no fast-check property test. Add tests/opencode-review-reconstruction.property.test.cjs. It extracts the two shipped jq programs (OPENCODE_REVIEW, OPENCODE_DIAG) verbatim from gsd-core/workflows/review.md and runs the real jq — not a reimplementation — so the shipped logic is what gets tested. Properties: the reconstructed review equals the newline-join of every assistant text part (order preserved); a stream with no text part reconstructs to empty (drives the #1936 stub); null/absent text parts are dropped, never rendered as "null". Plus example-based coverage of the diagnostic edges the reviewer cited: missing .tokens.output and no step_finish degrade to "?"; non-JSON stdout makes jq fail rather than masquerade as a review. Verified the invariant has teeth (a comma-join jq fails the property). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(#1936): skip jq reconstruction property test when jq is absent The property test shells out to `jq`, which GitHub's windows-latest runners do not ship (macOS/Linux runners do). `execFileSync('jq')` therefore ENOENT-failed the whole file on `test (windows-latest, *)`. Probe `jq --version` at load and skip the suite when jq is not on PATH — the reconstruction logic is platform-independent, so the assertions still run in full on every jq-present runner (mirrors how golden-install-parity skips on win32). Verified: jq present → 7 pass; jq removed from PATH → 7 skipped, 0 fail. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(#1936): skip jq reconstruction property test on Windows, not just when jq is absent The prior guard skipped only when `jq` was absent from PATH — but the windows-latest runners DO ship jq, so the suite still ran there and failed with `jq: parse error: Invalid numeric literal` (confirmed from the CI job log). Root cause is Node's child_process argument quoting mangling the jq program (it embeds double quotes) on Windows, not the shipped review.md logic — the macOS/Linux legs pass. Gate the suite on `process.platform === 'win32'` (still also skipping when jq is absent), mirroring golden-install-parity's win32 skip. Logic is platform-independent and fully asserted on every macOS/Linux CI leg. Verified: macOS → 7 pass; simulated win32 → skips. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-07-05 14:07:07 -04:00