agentleFS
SakuraByteCoreSakuraByteCore/codexmate/skills/context-recovery/SKILL.md

context-recovery

Skill356 starsChanged 9 months agoApache-2.0View on GitHub

Recovers project handoff context from local Codex, Claude Code, Gemini, CodeBuddy, and codexmate-derived sessions. Use when the user asks what happened in prior project/PR/branch/file/error work, needs a handoff brief, wants old decisions or validations recovered, or asks to summarize cross-session project activity with evidence.

3 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions3 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 3

    1. neomjs/neo-agent-skills★ 11same name
    2. omgpointless/aspy★ 2same name
    3. MarchRory/liushi_agent_settings★ 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.

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use plain search when you need raw candidates:

    line 22
  2. Very specific:

    Generate a brief: use --mode brief; add --path-filter for project/worktree isolation.

    line 32
  3. Distinctive:

    Live-check mutable facts: before commenting, merging, releasing, or claiming current status, check GitHub/current files directly.

    line 35
  4. Very specific:

    If codexmate MCP is configured and healthy, use it to inspect strong candidates:

    line 47
  5. Very specific:

    If MCP is unavailable, do not block.

    line 53

“Common” is measured: kept by at least 6 in 10 of the 3 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

Codexmate Project Context Recovery

Overview

Use this skill to build an evidence-first project handoff brief from local agent sessions. It is not semantic memory or a live GitHub source of truth. It works best with hard identifiers: owner/repo, branch names, file paths, commit hashes, exact errors, PR/issue numbers plus repo filters, or unique commands.

Old sessions are historical evidence. Mutable facts such as PR state, CI, releases, deployments, and current files still require live checks before action.

Quick Start

Generate a context brief first. Resolve scripts/search_sessions.py relative to this skill directory; after npm installation the full package path is node_modules/codexmate/skills/context-recovery/scripts/search_sessions.py.

python3 scripts/search_sessions.py "SakuraByteCore/codexmate feat/task-orchestration-tab" --mode brief --source all --path-filter codexmate --match all --format text --limit 8

Use plain search when you need raw candidates:

python3 scripts/search_sessions.py "exact error text" --mode search --source all --path-filter codexmate --format text --limit 10

Workflow

  1. Confirm the object: repo/project, PR/issue number, branch, file path, command, exact error text, person, or date range. If multiple projects match, state the chosen object before searching.
  2. Prefer hard identifiers: exact owner/repo, short repo name, PR/issue number with repo filter, branch, file path, commit hash, exact error, then user wording.
  3. Generate a brief: use --mode brief; add --path-filter for project/worktree isolation; use --match all when the query has strong identifiers.
  4. Check confidence: high means multiple hard signals appeared; medium/weak means re-query with stronger identifiers before relying on it; none means no hits were found.
  5. Use the brief as a handoff: extract timeline, decisions, validations, risks, files, commands, commits, and top evidence sessions.
  6. Live-check mutable facts: before commenting, merging, releasing, or claiming current status, check GitHub/current files directly.

What Good Output Looks Like

  • Target Object: repo / PR / branch / file / error
  • Context Brief: confidence, top sources, timeline, repos/branches/PRs/files
  • Historical Evidence: session source/id/path + snippets
  • Handoff Summary: decisions, validations, risks, commands, commits
  • Live Verification: PR/check/review/release/current-file facts that still need live verification

Optional codexmate MCP Path

If codexmate MCP is configured and healthy, use it to inspect strong candidates:

  • codexmate.session.list with source, query, queryScope: "all", limit, and forceRefresh: true.
  • codexmate.session.detail for candidate session inspection.
  • codexmate.session.export only when a markdown export is useful.

If MCP is unavailable, do not block; run scripts/search_sessions.py.

Limits

  • Generic natural-language queries can be noisy.
  • Short PR numbers without repo/path filters are weak signals.
  • Session logs may contain stale, failed, or speculative work.
  • The brief should guide investigation; it must not replace current repo/GitHub verification.

Privacy

Share only context relevant to the current task. Do not quote credentials, unrelated personal details, private memory, or large transcript chunks. In group chats, summarize narrowly and avoid exposing unrelated sessions.

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.