msd-core/commands/gsd/plan-review-convergence.md at a33cbe72f569e75e72d94de85d0930296d1ca1df

Files

Tom Boucher 81f9534b5a feat(adr-0002): command contract validation module + prose @-ref cleanup + workflow extraction

ADR-0002: commands/gsd/*.md contract now enforced at two layers:

LINT (scripts/lint-command-contract.cjs — new CI step):
- name: present, starts with gsd: or gsd-
- description: non-empty
- allowed-tools: non-empty, all entries canonical
- execution_context @-refs: resolve on disk, no trailing prose on same line
- handles both @~/ and $HOME/ path prefixes

TEST (tests/command-contract.test.cjs — 361 assertions):
- Behavioral contract for all 65 command files
- Replaces scattered coverage in enh-2790 + bug-3135
- Per-command per-rule test — one failure names the exact file + rule

CI (.github/workflows/test.yml):
- 'Lint — command contract (ADR-0002)' step added to lint-tests job

PROSE @-REF CLEANUP (39 command files, ~900 tokens/invocation recovered):
- Removed redundant @~/.claude/get-shit-done/... paths from <process> prose
- execution_context block is now the single authoritative load declaration
- Routing commands (sketch, spike, update, pause-work, etc.) keep routing
  instructions; only the inert path token is stripped

WORKFLOW EXTRACTION (debug.md + thread.md, ~15,000 chars / ~3,750 tokens):
- get-shit-done/workflows/debug.md: full process extracted from commands/gsd/debug.md
- get-shit-done/workflows/thread.md: full process extracted from commands/gsd/thread.md
- Command files reduced to frontmatter + objective + execution_context + context
- debug.md: 9,603 → 1,703 chars; thread.md: 7,868 → 585 chars

RENAME:
- get-shit-done/workflows/extract_learnings.md → extract-learnings.md
  (aligns with hyphen convention of all other workflow files)

DOCS:
- docs/INVENTORY.md: count 85→87, new rows, rename row, fix add-todo --backlog attribution
- docs/INVENTORY-MANIFEST.json: +debug.md +thread.md +extract-learnings.md -extract_learnings.md

Closes ADR-0002 implementation.

2026-05-05 15:18:13 -04:00

3.0 KiB

Raw Blame History

name, description, argument-hint, allowed-tools

name

description

argument-hint

allowed-tools

gsd:plan-review-convergence

Cross-AI plan convergence loop — replan with review feedback until no HIGH concerns remain.

<phase> [--codex] [--gemini] [--claude] [--opencode] [--ollama] [--lm-studio] [--llama-cpp] [--text] [--ws <name>] [--all] [--max-cycles N]

Read

Write

Bash

Glob

Grep

Agent

AskUserQuestion

Cross-AI plan convergence loop — an outer revision gate around gsd-review and gsd-planner. Repeatedly: review plans with external AI CLIs → if HIGH concerns found → replan with --reviews feedback → re-review. Stops when no HIGH concerns remain or max cycles reached.

Flow: Agent→Skill("gsd-plan-phase") → Agent→Skill("gsd-review") → check HIGHs → Agent→Skill("gsd-plan-phase --reviews") → Agent→Skill("gsd-review") → ... → Converge or escalate

Replaces gsd-plan-phase's internal gsd-plan-checker with external AI reviewers (codex, gemini, etc.). Each step runs inside an isolated Agent that calls the corresponding existing Skill — orchestrator only does loop control.

Orchestrator role: Parse arguments, validate phase, spawn Agents for existing Skills, check HIGHs, stall detection, escalation gate.

<execution_context> @$HOME/.claude/get-shit-done/workflows/plan-review-convergence.md @$HOME/.claude/get-shit-done/references/revision-loop.md @$HOME/.claude/get-shit-done/references/gates.md @$HOME/.claude/get-shit-done/references/agent-contracts.md </execution_context>

<runtime_note> Copilot (VS Code): Use vscode_askquestions wherever this workflow calls AskUserQuestion. They are equivalent — vscode_askquestions is the VS Code Copilot implementation of the same interactive question API. Do not skip questioning steps because AskUserQuestion appears unavailable; use vscode_askquestions instead. </runtime_note>

Phase number: extracted from $ARGUMENTS (required)

Flags:

--codex — Use Codex CLI as reviewer (default if no reviewer specified)
--gemini — Use Gemini CLI as reviewer
--claude — Use Claude CLI as reviewer (separate session)
--opencode — Use OpenCode as reviewer
--ollama — Use local Ollama server as reviewer (OpenAI-compatible, default host http://localhost:11434; configure model via review.models.ollama)
--lm-studio — Use local LM Studio server as reviewer (OpenAI-compatible, default host http://localhost:1234; configure model via review.models.lm_studio)
--llama-cpp — Use local llama.cpp server as reviewer (OpenAI-compatible, default host http://localhost:8080; configure model via review.models.llama_cpp)
--all — Use all available CLIs and running local model servers
--max-cycles N — Maximum replan→review cycles (default: 3)

Feature gate: This command requires workflow.plan_review_convergence=true. Enable with: gsd config-set workflow.plan_review_convergence true

Execute end-to-end. Preserve all workflow gates (pre-flight, revision loop, stall detection, escalation).

3.0 KiB Raw Blame History

3.0 KiB

Raw Blame History