map-maintain
Skill15 starsChanged 14 days agoApache-2.0View on GitHub
Run one bounded MYK stewardship cycle: map-check, Handler-selected repairs through map-this semantics, then targeted recheck. Trigger on "maintain this map", "tend this graph", or "/map-maintain". Not for autonomous monitoring.
In this repository95 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- agent-mailboxSkill
- agent-outputSkill
- ask-gateSkill
- ask-meSkill
- audit-coupled-constantSkill
- budget-checkSkill
- build-prototypeSkill
- caller-optionsSkill
- cold-reviewSkill
- debug-taskSkill
- define-termsSkill
- defuddleSkill
- design-an-interfaceSkill
- diff-recapSkill
- directives-auditSkill
- domain-checkSkill
- double-checkSkill
- extract-signalSkill
- freshness-contractSkill
- github-syncSkill
- handoff-executeSkill
- handoffSkill
- headless-claudeSkill
- headless-codexSkill
- headless-geminiSkill
- headless-grokSkill
- headless-modelsSkill
- headless-update-modelsSkill
- headless-usageSkill
- human-drawSkill
- human-mergeSkill
- human-outputSkill
- human-rewriteSkill
- implement-taskSkill
- improve-codebase-architectureSkill
- insight-adversarialSkill
- insight-anglesSkill
- insight-assessSkill
- insight-critiqueSkill
- insight-cross-examineSkill
- insight-exploreSkill
- insight-premortemSkill
- insight-reference-classSkill
- insight-retroSkill
- insight-root-causeSkill
- insight-skill-gapSkill
- insight-test-assumptionSkill
- insight-value-of-informationSkill
- investigateSkill
- json-canvasSkill
- map-checkSkill
- map-initSkill
- map-rulesSkill
- map-thisSkill
- monorepo-depsSkill
- multi-agent-modeSkill
- next-skillsSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
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
- Distinctive: line 23
Run one explicit, bounded stewardship cycle for an existing Map Your Knowledge (MYK) scope:
- Distinctive: line 71
Stop on overlapping agent activity, preimage drift, new ambiguity, or any required mutation outside the selected set.
- Very specific: line 94
Never say the graph is healthy without the original qualified map-check language.
≈ “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.
Files
/map-maintain
Run one explicit, bounded stewardship cycle for an existing Map Your Knowledge (MYK) scope:
map-check → findings → Handler selection → map-this apply semantics → targeted recheckThis skill is invoked work, never a daemon, schedule, watcher, background loop, or authority to repair future findings. It never reads, infers, records, or reports work-state. It does not replace map-check or map-this: it composes their accepted boundaries into one controlled maintenance pass.
Protocol provenance. MYK v2.4 comes from
map-rules/SKILL.mdand its routed references. Checking comes frommap-check/SKILL.mdand its required checker companion. Mutation semantics come frommap-this/SKILL.md.
Base kernel: load and obey map-rules. Use map-check as the single read-only conformance engine. Any selected mutation inherits the apply and environment rules in map-this; this skill grants no broader write authority.
Structured execution spec: protocol.yon. Read it for the canonical rules and step sequence; this file is explanation. The two must stay in sync — if you edit one, update the other and refresh the @STAMP date.
Preconditions and bounds
- Start only from an explicit invocation naming the target scope or path, from the Handler or from an agent whose current task plainly requires a maintenance cycle. Never auto-trigger from elapsed time, repository activity, a schedule, or a prior finding; the Handler selection gate before mutation remains unchanged.
- Resolve the exact governing scope and its contract before checking. Ambiguity, target drift, missing declaration, or a
map-checkoperational/resolution error stops the run without mutation. - Show the bounded run before execution: target, discovery mode, inventory source, file cap, finding cap, proposal page cap, elapsed-time cap, stop condition, and the Handler selection point. Defaults: at most 50 mutation-target files, 30 findings, 10 proposals per page, 30 minutes total, one repair-and-recheck round. Inventory and recheck walk the full governing scope using separately declared
--max-filesand--max-mschecker caps within that total budget; the 50-file mutation bound is not a scan bound. - Preserve every
map-checkqualification: checks run, not checked, exclusions honored, skipped boundaries, inventory completeness, mode, and discovery source. Never upgrade a partial or qualified result into an all-clear.
One maintenance cycle
1. Check
Run map-check read-only against the exact target. Stop with the qualified-clean verdict only when there are no in-scope actionable findings or normalized resolution occurrence cases. Zero conformance findings alone does not exclude actionable missing/ambiguous occurrences. Preserve confirmed finding and occurrence fields before bounded proposals; intentional red links or a conformance-only request do not manufacture repair work. If it returns an error envelope, stop and surface the remedy; absence of findings in an error is not evidence of health.
2. Prepare repair proposals
Translate confirmed findings AND the check's normalized resolution occurrence records/case files into simple map-this-shaped proposal rows — a missing or ambiguous link occurrence is a healing source even when it is not a conformance finding (an undeclared scope's broken links are occurrences, not C-findings). Every row stays case-bound to its exact finding or occurrence record; the no-creep rule is unchanged. Each row names the finding, exact files, exact mutation, why it repairs that finding, risk, uncertainty, and impact on maps, declarations, and links. Do not add cleanup, beautification, migration, lifecycle, archive, rename, or restructuring work merely because the scope is open.
Group coupled halves as one proposal: member ownership plus owner-map membership; archive metadata plus callout plus map relocation; rename/move plus every affected declaration and link. Preserve M3 exclusions and managed surfaces. A finding never authorizes its own repair.
Link-healing findings use the occurrence contract in link-resolution.md and the proposal classes owned by map-this — Link healing and enrichment proposals: create-alias-at-target, evidence-laddered retarget, disambiguate-ambiguous, rewrite-versus-typed-forwarding by measured fan-in, tag-and-defer, fingerprint-bound no-repair-needed, and inbound-ranked creation queues. Enrichment follows selected healing, targeted re-resolution, and graph rebuild as a separate report; without a stable resolved component it abstains. alias-chain-collapse may execute only as the mechanical half of an already Handler-selected exact set.
3. Handler gate
Offer accept, reject, modify, defer, or leave unresolved for every row. No mutation occurs until the Handler selects an exact set. Restate that set before applying it. Selection does not authorize expansion to neighboring findings.
4. Apply through map-this semantics
For the selected set only:
- Re-read every target preimage; stop and refresh if any changed.
- Resolve each write target and
lstat-check the leaf and parent chain; refuse symlinks, junctions, mount points, name-surrogate reparse points, and unknown reparse points. - Respect the resolved environment branch and all Lyt ownership boundaries. Never edit
.lyt/, use raw vault Git, broaden visibility, publish, delete, or mutate topology. - Apply coupled changes atomically where possible and idempotently always. On interruption, resume only a missing approved half after re-verifying the existing effect and unchanged target.
- A selected move/rename applies as ONE coupled change set with every selected referrer rewrite — preimage-checked, per-edit-annotated, resume-safe; healing writes inherit these rules unchanged and gain no new authority.
- Stop on overlapping agent activity, preimage drift, new ambiguity, or any required mutation outside the selected set.
5. Targeted recheck and stop
The current checker has no affected-only mode: rerun it on the full governing scope and record the actual command, inventory and finite caps. Limit interpretation to selected repairs and affected dependencies; reuse unrelated evidence rather than launching another audit. Budget exhaustion remains incomplete, never a manual review relabelled deterministic. Invalid-contract or incomplete resolution blocks dependent link healing/enrichment until contract repair and a complete recheck. Report repaired, remaining, deferred, and newly exposed findings separately; preserve the checker's qualifications and name checks not run.
The invocation ends after this one targeted recheck. It never chains into another repair round automatically. Remaining work needs a new explicit Handler invocation.
6. Enrichment suggestions (suggestion-only, separate report)
After the targeted recheck — and only when selected healing applied, re-resolution ran, fingerprints held, and the graph was rebuilt — generate link-enrichment SUGGESTIONS per map-this's owning section: bounded 2-hop candidates, filtered to the resolved connected component, evidence-laddered, ABSTAIN without a stable component. Suggestions express NEW intent, carry zero mutation authority, and are reported separately from repair; acting on any of them requires a new explicit Handler invocation. When this cycle's bounds are spent, hand enrichment off to a later bounded invocation by name rather than running it degraded.
Completion report
Return:
- exact scope, mode, and inventory boundary;
- original findings and Handler dispositions;
- changed paths and index-handoff result, if any;
- targeted recheck command and verdict;
- remaining, deferred, unresolved, or newly exposed findings;
- checks and mutation classes not run.
Never say the graph is healthy without the original qualified map-check language.
Human output. This skill's handler-facing output obeys the human-output contract (
human-output/SKILL.md).
Next skills. On completion, run the Next Skills protocol (
next-skills/SKILL.md): surface thenext-skillsrecommendations from front-matter for the caller to pick. Offer only — never auto-invoke.
Self-improvement. On completion, run the Self-Improvement Protocol (
self-improve/SKILL.md): apply its signal taxonomy and threshold to this run; surface any surviving proposal for handler approval. Conservative — silent otherwise. Never auto-apply.
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
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.

