agentleFS
QuirkybrainQuirkybrain/claude-code-swe-framework/.claude/skills/framework-checkpoint/SKILL.md

framework-checkpoint

Skill1 starsChanged 53 days agoMITView on GitHub

Persist a recoverable snapshot of project Stage, target, workflow, Artifacts, Components, decisions, assumptions, questions, risks, current path, evidence, and next action. Use at major results, pauses, blockers, escalations, integration, or session end.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    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
  1. Very specific:

    Run after a major Artifact/Gate result, before a blocking question or Target Stage pause, after escalation/replanning/integration, and …

    line 14
  2. Distinctive:

    Add recovery instructions and known State consistency limitations.

    line 38
  3. Very specific:

    Do not mark an Artifact VALID or Component STABLE without current evidence.

    line 44
  4. Distinctive:

    Stop when another session can identify exact state, evidence, blockers, and next safe action from persisted files, or when unresolved …

    line 74

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Framework Checkpoint

Purpose

Make project work resumable from versioned repository evidence rather than hidden chat or Agent memory.

Trigger

Run after a major Artifact/Gate result, before a blocking question or Target Stage pause, after escalation/replanning/integration, and before session end or expected context loss.

Inputs

  • Current State, Artifact Registry, Component states, active task, and Gate evidence.
  • Repository revision, branch/worktree identity, and working-tree status.
  • Current/Target Stage, Project/Task Workflow Levels, Execution Mode, and Current Minimum Necessary Path.
  • Decisions, assumptions, open questions, risks, blockers, and next ready action.

Prerequisites

  • Reconcile obvious State/repository discrepancies before declaring a clean snapshot.
  • Reference persisted Artifact paths and exact revisions rather than copying chat claims.

Procedure

  1. Assign a Checkpoint ID and timestamp/repository reference.
  2. Record repository revision, branch/worktree, dirty paths, and any uncommitted dependency relevant to recovery.
  3. Record Current Stage, Target Stage, Project Workflow Level, Task Workflow Level, and Execution Mode.
  4. Record Active Task, owner capability, status, dependencies, scope, and DoD.
  5. Record Artifact IDs/revisions/states and Component IDs/contracts/states with latest evidence.
  6. Record decisions, assumptions, Open Questions/Clarifications, risks, accepted deviations, and blockers.
  7. Record work completed since the previous Checkpoint and current validation/Gate verdicts.
  8. Record the Current Minimum Necessary Path, ready/blocked nodes, replan triggers, and next safe action.
  9. Add recovery instructions and known State consistency limitations.
  10. Validate that all referenced paths/revisions exist or mark them unresolved.

Decision Rules

  • Chat history and Agent memory are not authoritative recovery sources.
  • Do not mark an Artifact VALID or Component STABLE without current evidence.
  • Dirty working state must be explicit; do not imply it is committed.
  • Open material uncertainty remains an open clarification/assumption, not a decision.
  • A Checkpoint records state; it does not make multi-file updates atomic.

Outputs

A persisted Checkpoint containing at least:

  • Stage and Target
  • Project/Task Workflow Levels and Execution Mode
  • Artifacts and Components
  • Active Task and completed work
  • Decisions, Assumptions, Open Questions, Risks, and blockers
  • Validation/Gate evidence
  • Current Path, next action, and recovery instructions

Failure Cases

  • State files disagree or referenced revisions do not exist.
  • Working-tree changes cannot be attributed safely.
  • Evidence paths are missing.
  • Concurrent work makes the snapshot unstable.

Escalation

Mark conflicting entries UNVERIFIED and request reconciliation. Pause concurrent writers if necessary. Record Runtime-level atomicity limitations rather than claiming consistency guarantees.

Stop Conditions

Stop when another session can identify exact state, evidence, blockers, and next safe action from persisted files, or when unresolved inconsistency is explicitly recorded.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.