agentleFS
Kastalien-ResearchKastalien-Research/thoughtbox/.agents/skills/session-review/SKILL.md

session-review

Skill64 starsChanged 5 months agoMITView on GitHub

Generate a structured session summary for cross-session continuity. Captures key decisions, hypotheses, partial work, and knowledge references. Output is written to .Codex/session-handoff.json for the next session to load automatically.

10 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 6

    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
  1. Distinctive:

    Run git log --oneline -20 to see recent commits this session

    line 28
  2. Distinctive:

    Run git status to see uncommitted work

    line 29
  3. Distinctive:

    Run git diff --stat to see scope of changes

    line 30
  4. Very specific:

    Check .Codex/state/memory-calibration.json for pattern detection state

    line 31
  5. Distinctive:

    Read MEMORY.md for any updates made this session

    line 32
  6. Very specific:

    Write the session summary to .Codex/session-handoff.json (single file, overwritten each time) using this schema:

    line 47
  7. Distinctive:

    Verify the handoff file was written successfully

    line 108
  8. Distinctive:

    If there are uncommitted changes, warn about them

    line 110

“Common” is measured: kept by at least 6 in 10 of the 9 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.

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

Generate a structured session summary for cross-session continuity.

Context

$ARGUMENTS

Loop Building Blocks

LoopPurposeReference
Consistency CheckValidate session summary against evidence@loops/refinement/consistency-check.md
DocumentationGenerate structured session review@loops/authoring/documentation.md

See @loops/README.md for the full loop library.

Workflow

Phase 1: Gather Session State (Observe)

  1. Run git log --oneline -20 to see recent commits this session
  2. Run git status to see uncommitted work
  3. Run git diff --stat to see scope of changes
  4. Check .Codex/state/memory-calibration.json for pattern detection state
  5. Read MEMORY.md for any updates made this session

Phase 2: Synthesize Session Context (Orient)

From the gathered data, identify:

  1. Key Decisions: What architectural or design choices were made and why
  2. Open Hypotheses: What questions remain unanswered, what investigations are pending
  3. Partial Work: What was started but not finished, with enough context to resume
  4. Knowledge Discovered: New patterns, gotchas, or insights worth preserving
  5. Blocked Items: What's blocked and on what
  6. Next Steps: What the next session should prioritize

Phase 3: Write Session Handoff (Act)

Write the session summary to .Codex/session-handoff.json (single file, overwritten each time) using this schema:

{
  "version": "1.0.0",
  "session_id": "<from git>",
  "timestamp": "<ISO 8601>",
  "branch": "<current git branch>",
  "duration_estimate": "<approximate session duration>",
  "summary": "<1-2 sentence summary of what happened>",
  "key_decisions": [
    {
      "decision": "<what was decided>",
      "reasoning": "<why>",
      "alternatives_considered": ["<alt1>", "<alt2>"],
      "files_affected": ["<file1>", "<file2>"]
    }
  ],
  "hypotheses": [
    {
      "hypothesis": "<what's being investigated>",
      "evidence_for": ["<supporting evidence>"],
      "evidence_against": ["<counter evidence>"],
      "next_test": "<what to try next>"
    }
  ],
  "partial_work": [
    {
      "description": "<what was started>",
      "status": "<how far along>",
      "resume_from": "<specific file:line or commit to resume from>"
    }
  ],
  "knowledge_references": [
    {
      "store": "<MEMORY.md | Thoughtbox | git>",
      "reference": "<specific entity/issue/commit>",
      "relevance": "<why the next session needs this>"
    }
  ],
  "blocked_items": [
    {
      "item": "<what's blocked>",
      "blocker": "<what's blocking it>",
      "workaround": "<if any>"
    }
  ],
  "failed_approaches": [
    {
      "what": "<what was tried>",
      "why": "<why it failed>",
      "lesson": "<what to do instead>"
    }
  ],
  "next_priorities": ["<priority 1>", "<priority 2>", "<priority 3>"],
  "warnings": ["<things the next session should watch out for>"]
}

Phase 4: Verify and Report (Decide)

  1. Verify the handoff file was written successfully
  2. Print a human-readable summary to the console
  3. If there are uncommitted changes, warn about them

Output

Present a concise summary:

## Session Handoff Written

File: .Codex/session-handoff.json
Branch: {branch}
Commits this session: {N}
Open hypotheses: {N}
Partial work items: {N}

### Top 3 Priorities for Next Session
1. {priority}
2. {priority}
3. {priority}

### Warnings
- {warning}

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.