Audit a repo's agent context — CLAUDE.md files, codebase docs, skills, and tool/MCP designs — against Anthropic's Claude 5 context-engineering guidance ("unhobbling": Anthropic cut ~80% of Claude Code's system prompt with no eval loss). Finds overconstraint, conflicting instructions, redundancy, stale facts, and missing "unknown knowns"; produces a scored findings report with concrete rewrites, then applies approved fixes.
SkillScore82Maintained
Automatic
◆AI Engineer pack
Last verified 8/15/2026
What it changes
Passed 16 of 20 test cases with the skill, and 14 of 20 without it.
Without skill68%
With skill82%
Checks it passed that the model alone missed
Requests maintainer approval selecting which proposed edits to execute
Classifies the missing script reference as a high severity staleness finding
Converts strict formatting constraints into high-level guidance to match surrounding code style
and 2 more in the full report
5 fixed13 passed either way2 still failing2 broke
Pass-rate lift
+14 pp
Tokens
+89%
Agent turns
0%
Response time
-38%
Model gemini-3.6-flashMeasured 8/15/2026v118 of 22 cases passed with skill · 20 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
The audit is complete and the findings report contains 3 high severity findings, 5 medium severity findings, and 4 low severity findings. What is the mandatory next step before applying changes to the codebase?
What was checked, without and with the skill
Checked
Without skill
With skill
Requests maintainer approval selecting which proposed edits to execute
Missed
Passed
case-22 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md152 linessingle file
agent-context-audit — unhobble this repo's agent context
Goal: find where this repo's context (CLAUDE.md, docs, skills, tool designs)
hobbles a Claude 5-generation model — overconstrains it, contradicts itself,
repeats itself, or hides context the model actually needs — and leave behind a
findings report plus approved fixes.
Background: Anthropic removed over 80% of Claude Code's system prompt for
Claude 5 models with no measurable loss on coding evals. Older context was
written for models that needed rules; newer models need judgment, good
interfaces, and the facts they can't infer. This skill audits against that
shift, plus the "finding your unknowns" framework (the gap between the map —
your prompts/docs — and the territory — the actual codebase).
You are auditing first, fixing second. Do not edit anything until Step 4.
The six shifts (the audit rubric)
Every finding maps to one of these. Cite the shift number in the report.
Rules → Judgment. Hard rules ("NEVER…", "ALWAYS…", "do not add
comments", "one-line docstrings max") that encode a preference, not a real
constraint, should become judgment framing ("write code that reads like the
surrounding code") — or be deleted if the model would infer it anyway.
Keep hard rules only where violation is genuinely costly (security, prod
data, irreversible actions, legal/billing).
Examples → Interface design. Long tool-usage examples and few-shot
transcripts constrain exploration. Prefer expressive interfaces: good
parameter names, enums that hint at valid states, tight descriptions.
In tool/MCP definitions, an enum of pending | in_progress | completed
teaches more than three worked examples.
Upfront context → Progressive disclosure. Anything long that's only
sometimes needed (review checklists, deploy runbooks, style deep-dives)
should move out of CLAUDE.md into a skill or linked file loaded on demand.
CLAUDE.md is loaded every session — it should carry only what every
session needs.
Repetition → Concise, single-home instructions. The same instruction
appearing in CLAUDE.md and a skill and a tool description is a bug:
copies drift and eventually conflict. Each instruction gets exactly one
home — tool-usage guidance lives in the tool description, repo gotchas in
Audit the instructions an agent already carries — CLAUDE.md, AGENTS.md, skills, tool descriptions — for contradictions, over-constraint, and duplication, then propose a cut list. Use when an agent ignores its own instructions, when a CLAUDE.md has grown bloated, or when the user asks to audit or rightsize their agent context.