Conversation
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Source spec and its superpowers execution plan for Iterations 1 & 2 (already merged in 2b9149c). Commits the authoring artifacts so the rationale behind Rule 7, Step 0, and scope bounds is traceable in-repo. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Parses criterion ids (Rx.y) from a plan file and intersects them with completed_tasks[].covered[] from a progress.json to emit covered / uncovered / errors sets. Used by the new coverage-matrix-shape eval suite and available for future planner/verifier integration. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Planner now declares a risk_gates section at the top of every plan, enumerating tasks that trigger destructive ops, schema migrations, dep removals, or network state mutation. Each gate carries kind, reason, and confirmation prompt consumed by sea-go and executor in subsequent commits. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…nding marker Executor now checks the current task against the plan's risk_gates section before execution. On a gate match, writes .sea/phases/phase-N/gate-pending.json, marks progress.json task status 'gated', and exits with STATUS: gate. Resume path deletes the marker and proceeds as a normal task after sea-go captures user confirmation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… path Step 4.5 surfaces plan risk_gates entries for explicit user confirmation before executor launch. Step 5 grows a 'gate' case alongside done/blocked that reads gate-pending.json, collects user confirmation, and re-launches executor with resume context. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds the v2.1.0 gate-pending marker to the STATE.md inventory: writer, readers, required fields, invariants, and missing/corrupted behavior. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fixture plan exercising the v2.1.0 risk_gates taxonomy with one task per gate kind: schema-migration, dependency-removal, destructive-git, network-state-mutation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Structural simulation of the gate state machine: parses the fixture plan's risk_gates yaml block, asserts each expected gate kind is present with a non-empty confirmation, and round-trips a synthetic gate-pending.json marker through the fields sea-go reads on resume. Does not run a real executor. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds structural checks for v2.1.0 Iter 3: risk_gates in planner.md, Gate-pause protocol and STATUS: gate in executor.md, Risk gate inspection and Resume after gate in sea-go/SKILL.md. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Logs risk_gates schema, gate-pause protocol, Step 4.5/Resume-after-gate, gate-pending.json, fixture, risk-gate-flow suite, and prompt-quality extension. Flags pending live end-to-end validation as merge gate. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Merges feat/prompt-quality-risk-gates into develop. Iteration 3 installs the risk_gates taxonomy in planner, a gate-pause protocol in executor (STATUS: gate + gate-pending.json), and Step 4.5 risk gate inspection plus resume-after-gate handling in sea-go. Live end-to-end validation still pending per spec — see CHANGELOG 'Pending (Iter 3)'. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
demwick
added a commit
that referenced
this pull request
Apr 17, 2026
…guard Fix #1: Executor must complete all tasks or report STATUS: blocked. Sea-go now validates progress.json after STATUS: done and re-launches if tasks are missing. Ambiguous/missing STATUS treated as blocked. Sea-go never completes remaining tasks itself — only executor writes code. Fix #2: Sea-go Step 7d no longer blindly sets current_phase=N+1. When N >= total_phases, current_phase stays at total_phases and completed=true is set instead. Prevents out-of-range phase numbers. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Ships v2.1.0 prompt-quality layer across three iterations, all merged into
develop:_common.md, Step 0 (Demonstrate Comprehension) inresearcher.md,planner.md,executor.md.verifier.mdintentionally excluded.Allowed paths/Forbidden pathsin planner Mode B schema, pre-commit scope check in executor withscope violationblocked status,sample-plan-with-scope.mdfixture,scope-creep-detection.shsuite.risk_gatestaxonomy in planner, gate-pause protocol in executor (STATUS: gate+.sea/phases/phase-N/gate-pending.jsonmarker), Step 4.5 "Risk gate inspection" + "Resume after gate" insea-go/SKILL.md,sample-plan-with-gates.mdfixture,risk-gate-flow.shsuite,docs/STATE.mdmarker entry.scripts/check-coverage.sh+coverage-matrix-shape.sheval suite;.DS_Storeadded to.gitignore; source spec + superpowers plan committed underdocs/.Eval status:
bash evals/run.sh→ 23 passed, 0 failed.Test plan
bash evals/run.sh→ 23/23 greenbash evals/suites/agents/prompt-quality.sh→ passedbash evals/suites/agents/scope-creep-detection.sh→ passedbash evals/suites/agents/risk-gate-flow.sh→ passedclaude --plugin-dir "$(pwd)"against a throwaway repo withsample-plan-with-gates.mdas the target plan; observe executor pausing at gate task, sea-go surfacing confirmation prompt, and resume path completing. Log run in PR comment.🤖 Generated with Claude Code