plan-phases
Skill15 starsChanged 14 days agoApache-2.0View on GitHub
Restructure an existing plan into phases with /verify gates after each phase and critical outcomes, plus blocking pre-action gates for safety prerequisites. Trigger when the user runs /phase-plan or asks to "add phases to this plan", "gate this plan", "restructure the plan", or "add checkpoints". Use plan-create for new plans and plan-deep-dive for inspection without restructuring.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- Takkuya/claude-skillstheir own version, first seen Oct 20, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- Takkuya/claude-skills★ 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 repository95 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- agent-mailboxSkill
- agent-outputSkill
- ask-gateSkill
- ask-meSkill
- audit-coupled-constantSkill
- budget-checkSkill
- build-prototypeSkill
- caller-optionsSkill
- cold-reviewSkill
- debug-taskSkill
- define-termsSkill
- defuddleSkill
- design-an-interfaceSkill
- diff-recapSkill
- directives-auditSkill
- domain-checkSkill
- double-checkSkill
- extract-signalSkill
- freshness-contractSkill
- github-syncSkill
- handoff-executeSkill
- handoffSkill
- headless-claudeSkill
- headless-codexSkill
- headless-geminiSkill
- headless-grokSkill
- headless-modelsSkill
- headless-update-modelsSkill
- headless-usageSkill
- human-drawSkill
- human-mergeSkill
- human-outputSkill
- human-rewriteSkill
- implement-taskSkill
- improve-codebase-architectureSkill
- insight-adversarialSkill
- insight-anglesSkill
- insight-assessSkill
- insight-critiqueSkill
- insight-cross-examineSkill
- insight-exploreSkill
- insight-premortemSkill
- insight-reference-classSkill
- insight-retroSkill
- insight-root-causeSkill
- insight-skill-gapSkill
- insight-test-assumptionSkill
- insight-value-of-informationSkill
- investigateSkill
- json-canvasSkill
- map-checkSkill
- map-initSkill
- map-maintainSkill
- map-rulesSkill
- map-thisSkill
- monorepo-depsSkill
- multi-agent-modeSkill
- The first 60 shown.
The file
Show its rules and text
We split this skill into the rules it gives an agent
16 rules. About 0 are common, 8 are distinctive, and 8 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 35
Verify each condition is met
- Distinctive: line 36
Only continue to the next phase/step once the gate passes
- Very specific: line 38
If /verify is unavailable, verify the same conditions directly with available evidence.
- Distinctive: line 47
If no plan is found: after checking all locations → stop.
- Distinctive: line 58
Ensure phases flow logically — dependencies before dependents.
- Very specific: line 85
MUST insert a /verify gate after completing any phase
- Very specific: line 86
MUST insert a blocking pre-action /verify gate when authorization, no-follow target proof, backup or readback readiness, live-data scope, …
- Very specific: line 87
MUST retain a separate post-action /verify gate for the operation's outcome.
- Very specific: line 88
MUST insert an inline post-action /verify gate after any schema migration, breaking change, or security-sensitive step
- Very specific: line 89
MUST insert an inline /verify gate after any step with cross-package impact
- Very specific: line 90
SHOULD split phases with >5 steps into sub-phases with intermediate /verify gates
- Distinctive: line 100
MUST preserve ALL original plan content — restructure, never delete
- Distinctive: line 101
MUST maintain original intent and goal alignment
- Distinctive: line 102
MUST use the same file format as the original plan
- Distinctive: line 103
MUST NOT add new scope or features not in the original plan
- Very specific: line 104
MUST NOT create ceremonial /verify gates with vague focus
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/phase-plan
Take an existing flat or unstructured plan and restructure it into phased execution. Add /verify gates after each phase and critical outcome. Add blocking gates before actions whose safety prerequisites must already be true.
Structured execution spec:
protocol.yon. Read it for the canonical rules and step sequence; this file is explanation. The two must stay in sync — if you edit one, update the other and refresh the@STAMPdate.
Plan Restructuring Protocol — transforms any plan into a gated, phased structure for structural legibility and explicit checkpointing. Restructure, never delete. All original content is preserved.
What is /verify?
/verify is a verification gate — a point where execution stops and the current state is validated before proceeding. When you encounter a /verify gate in a restructured plan:
- Review what the gate specifies
- Verify each condition is met
- Only continue to the next phase/step once the gate passes
If /verify is unavailable, verify the same conditions directly with available evidence. User confirmation is evidence only for a Handler-owned decision, never a substitute for objective prerequisites. If a condition cannot be checked, keep the gate unresolved and stop its dependent action.
A /verify gate may sit before or after an action. It must block before mutation when it proves a prerequisite whose absence would make the action unsafe. Examples include authorization, no-follow target resolution, backup or readback readiness, live-data scope, and destructive-target confirmation. Retain a separate gate after the action to verify the intended outcome.
Phase 1 — Locate & Load Existing Plan
- Find the plan — when an explicit plan is supplied, select it; if the named source is missing or ambiguous, stop and ask for clarification. Otherwise check
implementation_plan.md,PLAN.md,task.md, or recent conversation context. Without an explicit source, resolve candidates against the current objective and context. If multiple plausible plans remain, surface the inferred target before replacement or return the proposed rewrite inline; modification time alone does not select a plan. Load the full content. - Assess structure — flat (single list)? Partially phased? Missing verification steps? Identify structural gaps.
If no plan is found after checking all locations → stop. Ask the user to provide the plan before continuing.
Phase 2 — Decompose Into Phases
- Identify natural phase boundaries — group related steps by dependency order, component, or concern. Each phase must be independently verifiable. Mark critical steps that need individual
/verifygates. - Define phases — each phase needs:
- Clear name
- Goal / deliverable
- Ordered steps
- Success criteria
Ensure phases flow logically — dependencies before dependents.
Phase 3 — Insert Check Gates
- Phase-level gates — insert a
/verifygate after every phase. Gate focus must be specific to what that phase accomplished and what the next phase depends on. Critical step gates — identify steps within phases that warrant individual
/verifygates:- Risky operations
- Breaking changes
- Schema migrations
- Cross-package impacts
- Security-sensitive changes
For each critical step, place any safety-prerequisite
/verifygate before the step, then place an outcome/verifygate immediately after it.
Gate Quality Reference
A gate must validate something observable. If it can't fail, it's ceremonial.
| Example | |
| Good gate | /verify — migration applied: users.email column is non-null, existing rows have values, no FK errors in test suite |
| Bad gate | /verify — make sure the migration looks right |
The bad gate has no observable condition. The good gate has three specific pass/fail criteria a different person could verify independently.
Gate Placement Rules
- MUST insert a
/verifygate after completing any phase - MUST insert a blocking pre-action
/verifygate when authorization, no-follow target proof, backup or readback readiness, live-data scope, destructive-target confirmation, or another safety prerequisite must be established before mutation - MUST retain a separate post-action
/verifygate for the operation's outcome; a safety prerequisite cannot be validated only after the action - MUST insert an inline post-action
/verifygate after any schema migration, breaking change, or security-sensitive step - MUST insert an inline
/verifygate after any step with cross-package impact - SHOULD split phases with >5 steps into sub-phases with intermediate
/verifygates
Phase 4 — Rewrite & Output
- Validate the candidate before replacement — check the complete proposed rewrite against original intent, every condition and all preserved content before saving. Before saving, validate the complete candidate, preserve the original bytes and unrelated or unknown content, and bind the exact target and current revision. Recheck identity immediately before a protected replacement; concurrent changes stop the write. Keep a recoverable preimage and read back persisted bytes. New-file creation is exclusive and must not replace an existing artifact.
- Rewrite the plan — in its original format (
implementation_plan.md,PLAN.md, ortask.md) with the new phased structure. Preserve all original content — restructure, do not delete. Each phase should clearly show: steps,/verifygate, and what the gate validates. If the plan has no source file (existed only in conversation context), output the full restructured plan as a fenced code block in the response instead. - Diff report — summarize what changed: how many phases created, how many
/verifygates added, any steps reordered or split.
Rules
- MUST preserve ALL original plan content — restructure, never delete
- MUST maintain original intent and goal alignment
- MUST use the same file format as the original plan
- MUST NOT add new scope or features not in the original plan
- MUST NOT create ceremonial
/verifygates with vague focus
Next Steps
/plan-execute— implement the newly phased plan/double-check— verify the restructured plan before executing
Human output. This skill's handler-facing output obeys the human-output contract (
human-output/SKILL.md).
Next skills. On completion, run the Next Skills protocol (
next-skills/SKILL.md): surface thenext-skillsrecommendations from front-matter for the caller to pick. Offer only — never auto-invoke.
Self-improvement. On completion, run the Self-Improvement Protocol (
self-improve/SKILL.md): apply its signal taxonomy and threshold to this run; surface any surviving proposal for handler approval. Conservative — silent otherwise. Never auto-apply.
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.

