Skip to content

JPeetz/agent-skills / skills/development/finding-unknowns/skills/context-audit

context-audit

Verified

Use when agent context is bloated or contradictory.

FileKindSize
CHANGELOG.mdmarkdown393 B
SKILL.mdskill md3.1 KB
evals/evals.jsonconfig2.5 KB

evals/evals.json

View on GitHub
{
  "skill": "context-audit",
  "version": "1.0.0",
  "cases": [
    {
      "id": "e1",
      "description": "Should trigger: bloated or contradictory agent context",
      "input": "My CLAUDE.md has grown to 200 lines and sometimes the agent contradicts itself. Audit it.",
      "shouldTrigger": true,
      "expect": "Inventories every context layer, classifies instructions, proposes a cut diff",
      "assertions": [
        "Inventories all layers (CLAUDE.md/AGENTS.md, skills, hooks, MCP descriptions) with sizes",
        "Classifies each instruction as conflict/duplicate/obvious/judgement-now/gotcha",
        "Quotes both sides of conflicts verbatim",
        "Proposes cut as a diff, conflicts first",
        "Ends with before/after line counts and the least-confident deletion named explicitly"
      ]
    },
    {
      "id": "e2",
      "description": "Should trigger: duplicate guidance across layers",
      "input": "We have the same rule in three files. Find the duplication.",
      "shouldTrigger": true,
      "expect": "Finds duplicates and proposes keeping the copy nearest point of use",
      "assertions": [
        "Reads layers together the way the model receives them",
        "Proposes relocation into on-demand files where appropriate"
      ]
    },
    {
      "id": "n1",
      "description": "Not trigger: adding context, not auditing it",
      "input": "Add a rule to CLAUDE.md saying we always use tabs.",
      "shouldTrigger": false,
      "expect": "Makes the edit directly",
      "assertions": [
        "No audit workflow, since this is an addition not a review"
      ]
    },
    {
      "id": "n2",
      "description": "Not trigger: unrelated question",
      "input": "What's 12 * 8?",
      "shouldTrigger": false,
      "expect": "Answers directly",
      "assertions": [
        "No context inventory performed"
      ]
    },
    {
      "id": "e3",
      "description": "Should trigger: cross-layer contradiction scan",
      "input": "Check whether our MCP server descriptions and the global preamble contradict each other.",
      "shouldTrigger": true,
      "expect": "Reads layers together, classifies conflicts, quotes both sides verbatim",
      "assertions": [
        "Reads layers together the way the model receives them",
        "Classifies instructions as conflict/duplicate/obvious/judgement-now/gotcha",
        "Quotes both sides of conflicts verbatim and prioritizes them",
        "Proposes a cut diff, doesn't apply it"
      ]
    }
  ]
}
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall