agentleFS
douincdouinc/agent-skills/skills/graph-engineering/SKILL.md

graph-engineering

Skill1 starsChanged 5 months agoMITView on GitHub

Use when work spans multiple components, an unfamiliar codebase needs dependency mapping, impact analysis depends on architecture or constraints, or durable evidence-backed project context must be created or refreshed.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 2

    Distinctive

    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:

    Do not require a graph database, MCP server, external API, or special runtime.

    line 22
  2. Distinctive:

    Default to read-only analysis.

    line 45

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “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

Graph Engineering

Overview

Model a project as evidence-backed knowledge, not a pile of files. Build only the graph needed for the current decision, keep facts separate from inference, and persist only durable knowledge.

Core Contract

  1. Define the question and the smallest useful scope.
  2. Inspect existing documentation before source code, tests, configuration, and history.
  3. Attach a repository path, symbol, sanitized command result, or commit to every important claim.
  4. Classify each claim as observed, inferred, proposed, or unknown.
  5. Represent relevant entities and connections with the vocabulary in knowledge-schema.md.
  6. Apply update-policy.md before changing durable documentation.
  7. Report impact, risk, unresolved questions, and the evidence that supports each conclusion.

Do not require a graph database, MCP server, external API, or special runtime. A compact Markdown table or diagram is enough when it preserves the required semantics.

Output Contract

Return these sections in order:

  1. Scope — question, boundaries, and audited revision when available.
  2. Evidence — strongest sources and any contradictions.
  3. Knowledge map — relevant nodes and directed relations.
  4. Impact — affected components, interfaces, constraints, and decisions.
  5. Risks and unknowns — consequences, missing evidence, and validation needed.
  6. Persistence decision — no change, update an existing document, or propose a new durable location.

Example map:

SourceRelationTargetStatusEvidence
AccountDeletionRouteimplemented_byDeleteAccountServiceobservedapi/routes/account.ts#deleteAccount
DeleteAccountServicedepends_onDeletionJobproposedapproved ADR-0042
DeletionJobconstrained_byIdempotent cleanupinferredretry configuration + integration test

Durable Knowledge Guardrail

Default to read-only analysis. Edit project knowledge only when the user explicitly requests documentation changes or the authorized task clearly includes them.

Never persist:

  • chain-of-thought, hidden reasoning, scratch notes, or loop transcripts;
  • raw command output or logs; persist a sanitized conclusion and command reference instead;
  • unverified guesses presented as facts;
  • secrets, credentials, personal data, or environment-specific identifiers;
  • temporary status that will be obsolete when the current task ends.

Common Mistakes

MistakeCorrection
Mapping every fileKeep only entities that affect the current decision.
Treating a diagram as evidenceAttach evidence to each important node or relation.
Replacing conflicting historyMark the conflict and supersede only with stronger evidence.
Creating a new context folder immediatelyPrefer the closest existing durable document.
Writing assumptions as architectureLabel them inferred, proposed, or unknown.

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.