agentleFS
ErionisErionis/Jarvis-Lite/skills/jarvis-doctor/SKILL.md

jarvis-doctor

Skill3 starsChanged 3 months agoMITView on GitHub

Use when an existing Jarvis appears inconsistent or needs a local safety check.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. Hainrixz/jarvis-architect★ 18same 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.

  • In this repository10 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not use starter paths, filesystem discovery, or guessed fallbacks.

    line 30
  2. Very specific:

    Treat <!-- jarvis:onboarding-required --> in the resolved local profile only as the onboarding marker.

    line 35
  3. Very specific:

    Use an available read-only effective-access check or permission metadata.

    line 75
  4. Very specific:

    Use exactly one label: error, unverifiable, or optional evolution.

    line 131
  5. Very specific:

    After the report, actionable findings may justify a separate plan.

    line 146

≈ “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

Jarvis Doctor

Audit contract, readiness, and consistency from local evidence. Diagnose only: jarvis-memory remains the only curator of Identity and Durable memory.

Scope and focus

Scope is the profile and bootstrap contract, guardrails, declared sources, and one direct local-reference hop. Do not use remote services as evidence or scan the workspace. Preserve local extensions outside the common contract. Do not normalize or remove them.

A plain Doctor request audits the complete bounded scope. If the user names a symptom or capability, report that role and its direct dependencies. Include bootstrap blockers, never unrelated bootstrap-only findings. Do not introduce flags or a second mode.

Doctor is read-only. Do not edit, create, rename, move, delete, or install anything. Do not generate a report file. Do not stage, commit, push, or configure Git; never perform repair as a side effect.

1. Resolve the contract

  1. Resolve every capability role from the consumer's declared capability map. Do not use starter paths, filesystem discovery, or guessed fallbacks. Identify each declaration as unique, missing or ambiguous. Check whether each declared target exists and is readable. A declared missing target is evidence; do not silently create it. If a prerequisite is unreachable, report only what direct evidence proves.
  2. Treat <!-- jarvis:onboarding-required --> in the resolved local profile only as the onboarding marker. A Doctor request is concrete user work: complete the bounded read-only diagnosis before offering setup. When Identity is missing or the onboarding marker is present, do not bypass first run: report Bootstrap blocked and route to first-run. While first run is routed, Git-pending markers belong to that route; do not report them separately. A missing non-Identity capability is unavailable and does not authorize a fallback source. A missing optional capability is not automatically an error; an invalid declared target is.
  3. Once Identity exists and the onboarding marker is absent, count exact <!-- jarvis:git-pending --> markers in the resolved local profile. The marker is restore-point state owned by first-run, not a bootstrap blocker.

    • Exactly one marker: the local restore point is pending. Say so in plain language as optional evolution: ordinary work continues, and the user can ask Jarvis to enable local restore points. Lead with that state, keep the marker as technical evidence, and route the action to the restore-point question in first-run.
    • Zero markers and no Git repository: restore points were declined or never set up. Report it as one neutral line within the inspected scope, with no label, action, or reminder.
    • Zero markers in a Git repository: no restore-point finding.
    • More than one marker: report the exact count as ambiguous restore-point state, labeled error. Recovery belongs to first-run, which asks for approval before reducing it.

    Establish repository presence only with read-only evidence, such as git rev-parse --is-inside-work-tree. Missing Git counts as no repository and is not a finding. Do not ask the restore-point question, add or remove a marker, or initialize Git.

2. Verify structure and readiness

Common Lite roles use Markdown files for Identity, Soul template, Durable memory, and Future work, and directories for Daily history, Handoff, and Inbox. Do not impose these shapes on local roles.

A full audit checks every declared writer: Identity, Durable memory, Future work, Daily history, Handoff, and Inbox. A focused audit checks only the named writer and its dependencies. For each, assess writability without mutation. Use an available read-only effective-access check or permission metadata; certain denial is a verified error. If the runtime cannot establish it without mutation, readiness is unverifiable. Do not create a write probe, script, cache, or temporary file in the consumer.

The consumer's established Handoff contract is the lifecycle authority. Use the already-discovered handoff skill and bootstrap contract; do not search for another schema. The shipped Lite contract defines records as Markdown files whose frontmatter contains type: handoff; valid states are active, completed, and superseded. A missing or unknown status is a verified error. created: and updated: are required on every Handoff record. completed requires completed:; superseded requires superseded: and superseded_by:. resumed: is metadata, not a status. Do not lint ordinary Markdown files as handoffs or change records. Preserve lifecycle extensions declared by an adapted consumer contract.

Daily history is historical evidence. Open it beyond readiness only when an active source directly cites it for a current fact.

3. Inspect bounded semantics

  • Profile: A repeated shared rule is optional evolution; an incompatible directive or expired override is an error. Bootstrap reminders, capability declarations, and local context are not duplicates. Do not judge style or length.
  • Current facts: Require an explicit pointer, time-bound promise, or directly verifiable filesystem evidence. A past date alone is not stale. Mark a current-state claim unverifiable only when its own date or freshness statement no longer supports present-tense use, or the declared evidence cannot establish it. Do not invent a universal age threshold. Do not turn preferences or descriptive present-tense prose into findings.
  • Contradictions: Require two mutually incompatible claims about the same subject with compatible scope and time. Different dates or scopes are not automatically contradictions. If scope cannot be resolved, label the finding unverifiable, not error.
  • Authority: A normative directive duplicated across Identity and Durable memory is optional evolution; incompatible directives are errors. Pointers, examples, and history are not duplication. Consolidation or repair is a separate jarvis-memory request with a preview.
  • Local links: Ignore external URLs and fragment-only anchors. Resolve each local relative Markdown link relative to its declaring source. Ignore links inside code. For a missing target, retain the literal link and resolved missing target as evidence.

Group equivalent low-impact findings; keep bootstrap blockers and failures in different capabilities separate.

4. Report operationally

Open in the user's language with the equivalent of one state: Jarvis operational, Jarvis partially operational, or Bootstrap blocked, and say that nothing was changed. Use partial status for non-bootstrap errors or unverifiable readiness. Other uncertainty and optional evolution do not degrade status.

Use exactly one label: error, unverifiable, or optional evolution.

  • error: a mechanically verified contract or readiness failure.
  • unverifiable: evidence cannot support a reliable conclusion.
  • optional evolution: a non-required improvement, never a disguised failure.

State the inspected scope. For each finding, include a compact source path and location when applicable. Explain problem, impact, evidence, and minimal repair in plain language before technical detail. End with at most three prioritized actions routed to the owner. With no findings, about three lines are enough.

Report inspected scope and count findings only after verifying what was read. With no findings, name which roles, sources, and link scope were checked. Do not claim the whole installation is healthy beyond that evidence.

After the report, actionable findings may justify a separate plan. Offer to update an existing plan or ask where a new one belongs; if Future work exists, offer one pointer. A plan or repair is separate follow-up work. Never create or update it during Doctor. Without actionable findings, do not offer a plan.

Doctor does not certify external services or preserve audit state.

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.