phase-gate
Skill0 starsChanged 4 months agoMITView on GitHub
Use when closing a project phase or milestone gate — the user believes the phase's acceptance criteria are met and wants it closed formally, with evidence verified item by item. Runs the closure ritual in strict order and blocks on any open item. Trigger phrases: "close the phase", "close the gate", "is the phase done", "phase-gate check", "milestone closure", "can we call this phase complete".
5 other projects use a version of this.
Adopted by
Most recently taken up by
- acprk/crypto-research-agentstheir own version, first seen Nov 4, 2026
- andreymudri/claude-teammatestheir own version, first seen Nov 2, 2026
- andreymudri/fleetmatestheir own version, first seen Nov 2, 2026
- Ahsan-ire/rag-pipeline-v1their own version, first seen Oct 12, 2026
- acgs-ai/ACGStheir own version, first seen Aug 6, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- acprk/crypto-research-agents★ 15same name
- andreymudri/claude-teammates★ 3same name
- andreymudri/fleetmates★ 3same name
- acgs-ai/ACGS★ 2same name
- Ahsan-ire/rag-pipeline-v1★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository15 other files it gives its agents.
- templates/CLAUDE.mdCLAUDE.md
- build-logSkill
- canonical-paramsSkill
- correctness-gateSkill
- decision-logSkill
- evidence-conventionSkill
- layer-syncSkill
- project-layoutSkill
- run-provenanceSkill
- external-positioningSkill
- lit-anchorSkill
- negative-results-ledgerSkill
- numerical-determinismSkill
- statistical-reportingSkill
- skill-library-generatorSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 28
Load the contract: Read the phase's acceptance checklist from {{phase_plan}} (and its gate registry row, if the plan keeps one).
- Very specific: line 36
Decision sweep: Every decision this phase owed (check the plan's decision queue, if any) must exist in {{decision_log}} with a real number …
“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
phase-gate
Close a phase the way a reviewer would: verify the phase's written acceptance checklist item by item against named evidence artifacts, then update the record everywhere it lives. A phase closes on evidence, not on recollection — and one open item means it does not close.
When to use
- A phase, milestone, or gate from the project plan (
{{phase_plan}}) appears complete and the user wants formal closure.
When NOT to use
- Mid-phase "is this sub-task done?" — that is an execution-turn question against the current TODO buffer.
- "Where are we / what's next" — dispatch the
session-historianagent. - Declaring partial victory — there is no partial closure (see Rules).
The ritual (strict order — no step skipped, no step reordered)
- Load the contract. Read the phase's acceptance checklist from
{{phase_plan}}(and its gate registry row, if the plan keeps one). This written checklist — never memory — is what gets verified. - Verify each item with evidence. For every checklist item, name the artifact that
proves it: fresh gate-test output, an
{{evidence_dir}}/<name>_<date>/run directory with its meta file, a decision entry, an executed notebook or committed figure. An item without an artifact is OPEN — the phase does not close. Run{{gate_command}}fresh as part of this step; a red gate blocks closure regardless of checklist state. - Decision sweep. Every decision this phase owed (check the plan's decision queue, if
any) must exist in
{{decision_log}}with a real number — assigned at write time, never reserved in advance. - Build-log entry. Invoke the
build-logskill (phase-completion bar applies). The entry links every evidence artifact from step 2. - Layer-sync audit scoped to the phase. Drift found → fix in a separate approved turn per that skill's rules, then re-run the audit before closing.
- Update the status surfaces. Mark the phase ✅ on the plan's status board, and rewrite the TODO buffer to the next phase's sub-tasks — the buffer is ephemeral; completed-phase content is deleted, not accumulated.
- Regenerate derived views (project wiki, dashboards, generated indexes), if the project keeps any. Derived views are never hand-edited: wrong content means a source-layer fix first, then regeneration.
- User sign-off. Present the closure summary — checklist verdicts, evidence links, decision/build-log diffs — and get explicit approval before committing. Gate decisions with branches (pass/drop calls, representative-point choices) are the user's call, never auto-resolved.
Output format
## Phase closure — <phase name> / <gate id>
| Acceptance criterion | Evidence | Status |
|---|---|---|
| ... | {{evidence_dir}}/<run>/ or build-log NN or D<n> | ✅ / ❌ OPEN |
**Gate test**: <N> tests green (summary of fresh output)
**Decisions**: D<n>, ... | **Build log**: NN_topic.md | **Layer sync**: ✅/⚠
**Verdict**: CLOSED / BLOCKED — <open items>Rules
- One OPEN item = no closure. Pre-declared off-ramps written into the plan (e.g. "if the result window is empty, branch to X") are legitimate closures; silent scope-shrinking is not.
- Evidence means versioned evidence. "It ran and the output looked fine" without an
{{evidence_dir}}run directory is a dark run — treat the item as OPEN (seeevidence-convention). - Never soften a negative result to pass a gate. An infeasibility, a null effect, or a
hardness finding — rigorously recorded — is a valid closure artifact; record it
verbatim and close on it (see
negative-results-ledger). - Commits happen after sign-off, following the project's git workflow.
- Fresh runs only. Cached test output, yesterday's logs, or "it passed last week" do not satisfy step 2.
Configuration
{{phase_plan}}— the plan document holding phase tables + acceptance checklists.{{gate_command}}— the named correctness gate (seecorrectness-gate).{{decision_log}},{{build_log_dir}},{{evidence_dir}}— the record surfaces.
Provenance & maintenance
Generalized from the gate-closure ritual of a working computational-research repository
(checklist-with-evidence verification, decision sweep, build-log + sync audit coupling,
sign-off-before-commit); see the pack's examples/ directory for the worked exemplar
mapping. The eight-step order and rules 1–3 are verified practice; step 7 applies only if
the project keeps derived views (labeled optional by construction).
Re-verify in your project:
{{gate_command}}— must be green before any closure claim.grep -n "OPEN\|⬜\|🟡" <status board>— anything still open on the board contradicts a closure claim.ls -t {{evidence_dir}} | head -3— the runs cited as evidence actually exist and are recent.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

