Internal helper — parses a plan file's `## Convergence` section and evaluates its `criterion` shell command against current state. Returns a structured report with `met`, `evidence`, and `gap` fields. Called by /verify and /status; not for direct user invocation.
インストール
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
Internal helper — parses a plan file's `## Convergence` section and evaluates its `criterion` shell command against current state. Returns a structured report with `met`, `evidence`, and `gap` fields. Called by /verify and /status; not for direct user invocation.
when_to_use
Reach for this when /verify or /status needs to evaluate whether a declared convergence criterion is satisfied before claiming a sprint is done. Do NOT use for ad-hoc edits without a plan — convergence only applies when a plan declares it. Use /verify directly for plan-less work.
disable-model-invocation
true
allowed-tools
["Read","Bash"]
scheduling
an active plan declares a `## Convergence` section with a `criterion` field, and /verify or /status needs to evaluate it
structural
["Locate the plan file (argument or active plan resolved via `find-active-plan.sh`)","Parse the `## Convergence` section for `type`, `criterion`, and `max_iterations`","Execute the criterion command with a 10-second timeout","Report exit code, stdout evidence, and gap description"]
logical
stdout contains a structured report with `plan_path`, `convergence_type`, `criterion`, `criterion_exit_code`, `met`, and `evidence_lines`; exit 0 when criterion is met, 1 when unmet, 2 when no convergence block, 3 when plan not found
convergence-check — Evaluate a Plan's Convergence Criterion
Internal helper skill. Called by /verify (to gate "done" claims) and /status (to show progress vs criterion). The user rarely invokes this directly — it surfaces through the callers.
What it evaluates
A plan's ## Convergence section declares a machine-checkable predicate:
The criterion value is a shell command. This skill runs it with a 10-second timeout and checks the exit code: 0 = criterion met, non-zero = criterion unmet.
Trust model: the user wrote the plan file. Commands in criterion run with the same privilege as any other shell command the user authorizes. The user accepts this when they write the plan.
Execution Checklist
Locate the plan file — use the argument if given; otherwise resolve via plugins/workflow/skills/orchestrate/scripts/find-active-plan.sh (single source of truth for the active plan)
Verify the plan file exists — exit 3 if not found
Parse the ## Convergence section for type, criterion, and max_iterations — exit 2 if no convergence block
Execute criterion in a subshell with timeout 10 — capture stdout, stderr, and exit code
Emit a structured report to stdout: plan_path, convergence_type, criterion, criterion_exit_code, met (true/false), evidence_lines
Exit 0 if criterion met, 1 if unmet, 2 if no convergence block, 3 if plan not found
Optional plan-path argument: absolute or repo-relative path to a plan file. If omitted, the script resolves the active plan via plugins/workflow/skills/orchestrate/scripts/find-active-plan.sh (the canonical resolver — same one used by /orchestrate, /contract, and the after-subagent hook). If the resolved plan has no ## Convergence section, exit 2 (skip gracefully — implicit convergence is valid for one-shot edits).
