agentleFS
stevehansenstevehansen/eidet/.claude/skills/memorytool/SKILL.md

memorytool

Skill4 starsChanged 54 days agoMITView on GitHub

Prime on Eidet's memory_20250818 backend before changing it — MemoryToolTranslator's six commands (view/create/str_replace/insert/delete/rename), MemoryPath validation, byte-exact path-keyed blobs in memoryfiles/*, the size cap and SecretPolicy, the reserved /memories/.recall subtree, and the opt-in one-way IMemoryBridge into the semantic store. Use when the task touches Claude memory-tool commands, /memories paths, MemoryFile storage, or POST /api/eidet/memory-tool. Not for semantic memories (see memory), not for seeding from repo files (see intake).

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Nothing the model can do throws: Every expected failure is an is_error result with the contract's exact wording.

    line 17

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Memory tool — priming

Canonical spec: docs/domains/memorytool.md — read it for command behaviour, all invariants, key files, and gotchas. Terms of record: UBIQUITOUS_LANGUAGE.md § Memory-tool files. Threats: STRIDE.md I-8, T-15.

A faithful filesystem: the model re-reads exactly the bytes it wrote. A Memory file is not a Memory — it skips the signal gate, decay, and consolidation by design.

Core invariants (get these right)

  • Nothing the model can do throws. Every expected failure is an is_error result with the contract's exact wording; a storage fault returns a generic error and logs server-side (never leak ex.Message).
  • MemoryPath is the only path validator — traversal, backslashes, control chars, and url-encoded escapes are rejected at construction (I-8). Never concatenate a blob key.
  • The gate here is the secret scan + size cap only — the semantic low-signal/self-talk gates would break the filesystem contract. Secret scanning is still mandatory.
  • A redacting write reports its redaction count; a rejecting policy stores nothing.
  • Never rewrite bytes semantically — occurrence-counted str_replace, bounds-checked insert, verbatim store.
  • The blob is the source of truth; the bridge is best-effort, one-way, and off by default (NullMemoryBridge). Semantic-side rejections and duplicates are expected.
  • /memories/.recall is read-only and reserved; the root is never a file.
  • Repo isolation is bound at construction — one translator per repo.

Key files / reuse

  • src/Eidet.Core/MemoryTool/MemoryToolTranslator.cs — the single ExecuteAsync entry.
  • src/Eidet.Core/MemoryTool/MemoryPath.cs — parse/validate here, always.
  • src/Eidet.Core/MemoryTool/MemoryToolOptions.cs — size cap + SecretPolicy.
  • src/Eidet.Core/MemoryTool/InMemoryFileStore.cs — the test store.

Gotchas

  • Directories are implied by deeper blob keys, never stored — an empty directory can't exist.
  • Line handling is logical: the trailing-newline phantom line is dropped; view is 1-based with -1 meaning end-of-file.
  • The error strings are part of the contract and are pinned by tests — rewording changes model behaviour.
  • Promoted files become memory-tool-tagged Observations carrying the file path as a tag; .recall depends on that tag shape.
  • The size cap counts UTF-8 bytes, not characters.

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.