agentleFS
TheGoat395TheGoat395/Codex-Skills/skills/agent-memory-provenance/SKILL.md

agent-memory-provenance

Skill136 starsChanged 5 months agoMITView on GitHub

Design memory provenance and tenant isolation.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    run state: transient tool and workflow state.

    line 12
  2. Distinctive:

    Do not treat an LLM summary as evidence or a vector result as permission to act.

    line 18
  3. Distinctive:

    Treat stale or conflicting retrieval as a question to resolve, not silent context.

    line 30
  4. Distinctive:

    Keep client credentials, documents, logs, and decisions isolated by default.

    line 31

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Agent Memory & Provenance

Separate memory types

Keep these distinct:

  • run state: transient tool and workflow state;
  • session history: bounded conversational continuity;
  • evidence store: source documents, extracts, timestamps, and access status;
  • decisions: approved choices, assumptions, owner, and reconsideration condition;
  • retrieval index: derived search aid, never the sole record of truth.

Do not treat an LLM summary as evidence or a vector result as permission to act.

Design the record

For each retained item, define tenant/owner, source URI or file, capture time, fact-versus-inference label, allowed readers, retention period, deletion path, and sensitivity. Preserve originals when possible; make derived summaries reversible by linking them to their source records.

Use append-only non-sensitive event references for auditability; sensitive payloads need their own authorized deletion and retention path. Existing authorization governs storage; a memory design task does not authorize adding personal durable memories. Compact only with an explicit summary of what was retained, omitted, and still retrievable. Never place passwords, tokens, private keys, cookies, recovery codes, or unrelated client data in agent memory.

Retrieval and isolation

  • Retrieve only records relevant to the active task and permitted tenant.
  • Filter by source authority, freshness, locale, and client before semantic similarity.
  • Treat stale or conflicting retrieval as a question to resolve, not silent context.
  • Keep client credentials, documents, logs, and decisions isolated by default.

Verify and deliver

Test the changed behavior and material failure boundaries, selecting from cross-session recall, source citation, conflicting sources, sensitive-record exclusion, retention and tenant isolation. Deliver the requested result; select a memory map, data dictionary, retention matrix or evidence receipt only when needed for that result. Use $agent-orchestration-architecture for runtime ownership and $research-provenance-archive for durable research capture.

Optional specialists

For durable research archives, research-provenance-archive is optional. Without it, preserve attributable source identifiers, capture dates, claim/evidence links, access boundaries and corrections in the existing project store.

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.