agentleFS
Sign inStart free
modeled-information-formatmodeled-information-format/mnemonic/skills/custodian/SKILL.md

custodian

Skill23 starsChanged 10 months agoMITView on GitHub

Memory system custodian for health checks, validation, and maintenance. Trigger phrases: "check memory health", "validate memories", "fix broken links", "update decay", "relocate memories", "audit memories", "memory maintenance", "custodian", "memory health"

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    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:

    Run /mnemonic:list --namespaces to see available namespaces from loaded ontologies.

    line 23
  2. Distinctive:

    Proactively: when session_start reports low health scores or potential duplicates

    line 32
  3. Distinctive:

    On request: when user asks about memory health, broken links, or maintenance

    line 33
  4. Distinctive:

    Read-only validation. Reports errors without modifying files.

    line 125
  5. Distinctive:

    Atomic writes: All file modifications use temp file + replace pattern

    line 137
  6. Distinctive:

    Single-pass audit: All checks run once per file for efficiency

    line 140

≈ “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
<!-- BEGIN MNEMONIC PROTOCOL -->

Memory

Search first: /mnemonic:search {relevant_keywords} Capture after: /mnemonic:capture {namespace} "{title}"

Run /mnemonic:list --namespaces to see available namespaces from loaded ontologies.

<!-- END MNEMONIC PROTOCOL -->

Custodian Skill

Provides custodial services for the mnemonic memory system: auditing, validation, link repair, decay management, relocation, and summarization.

When to Use

  • Proactively when session_start reports low health scores or potential duplicates
  • On request when user asks about memory health, broken links, or maintenance
  • After project renames to relocate memories and update references
  • Periodically to keep decay scores current and identify orphaned memories

Architecture

The custodian operates through focused Python modules:

ModuleResponsibility
memory_file.pyParse/validate/update MIF frontmatter
link_checker.pyBuild UUID/slug index, validate links, find orphans
decay.pyCalculate exponential/linear/step decay, update strength
relocator.pyMove files + update all cross-references
validators.pyMIF schema validation, ontology relationship checks
report.pyStructured findings with markdown/JSON output
custodian.pyCLI orchestrator dispatching to modules

Subcommands

audit (default)

Runs all checks in a single pass. This is the recommended entry point.

/mnemonic:custodian audit [--fix] [--dry-run]

Checks performed:

  1. Frontmatter: Required fields, UUID format, type enum, date format
  2. Links: Wiki-links resolve to existing memories
  3. Relationships: Types match MIF built-in or ontology-defined types
  4. Decay: Recalculates strength values using configured model
  5. Orphans: Detects memories with no incoming references

relocate

Moves memories when project or org names change:

/mnemonic:custodian relocate <old-path> <new-path> [--dry-run]
  1. Builds link index across all memory roots
  2. Updates path references in ALL memory files (not just relocated ones)
  3. Moves files using git mv for history preservation
  4. Cleans up empty directories

validate-links

/mnemonic:custodian validate-links [--fix]

With --fix: broken [[wiki-links]] are replaced with plain text.

decay

/mnemonic:custodian decay [--dry-run]

Updates temporal.decay.currentStrength (or strength) in-place. Formula: strength * 0.5^(days_since_access / half_life_days)

summarize

/mnemonic:custodian summarize

Identifies compression candidates (>100 lines, >30d old or strength < 0.3). Delegate actual compression to mnemonic:compression-worker agent.

ensure-bidirectional

/mnemonic:custodian ensure-bidirectional [--fix]

Scans all memory relationships and checks that each has a corresponding inverse back-reference in the target. For example, if A supersedes B, B should have superseded-by pointing back to A. Legacy spellings (SupersededBy, superseded_by) and bare-UUID targets count as present; new back-references are written in MIF 1.4.2 form (kebab-case type, urn:mif:<uuid> target, plus a body ## Relationships link).

Without --fix: reports missing back-references as warnings. With --fix: auto-creates the missing inverse relationships.

validate-memories / validate-relationships

Read-only validation. Reports errors without modifying files.

Integration

  • Complements gc: GC handles cleanup/archival; custodian handles validation/repair
  • Complements memory-curator: Curator handles conflicts/deduplication; custodian handles schema/links/decay
  • Uses PathResolver: Centralized path resolution from lib/paths.py
  • Uses compression-worker: Delegates summarization to existing agent

Key Design Decisions

  • Non-destructive by default: --dry-run support on all mutation operations
  • Atomic writes: All file modifications use temp file + replace pattern
  • Git integration: --commit flag for committing changes
  • Functional modules: Lightweight functions over heavy OOP, matching codebase style
  • Single-pass audit: All checks run once per file for efficiency

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.

Check your setup, free