Context Window Audit
Audit an agent setup for token waste, context-window bloat, stale instructions, and dead command wrappers. Use when the user says context-window audit, context audit, audit my context, check my settings, why is Codex so slow, token optimization, or asks why startup context is too large.
Install
$ agentstack add skill-mattdweigand-sketch-agent-skills-context-window-audit ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Links to your public security report.[](https://agentstack.voostack.com/security/report/skill-mattdweigand-sketch-agent-skills-context-window-audit) Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Context Window Audit
Use this skill when the user asks for a context-window audit, context audit, token optimization, settings cleanup, or an explanation of why an agent session is loading too much context.
Scope
Audit prompt and context surfaces that load automatically or are likely to be pulled into many sessions:
- global and repo instruction files such as
AGENTS.md,CLAUDE.md, command
wrappers, settings, and skill descriptions
- enabled skills, plugins, and connectors when visible
- memory summaries or routed context files when they affect startup behavior
- repeated references that cause large files to be read by default
Do not audit private content for substance unless it is relevant to context loading. The goal is smaller, clearer context, not content cleanup.
Load references/context-surface-taxonomy.md when classifying issues or formatting the final audit. That file owns the stable action taxonomy and output template.
Process
- Identify the startup path.
- List the files, skills, commands, memories, or settings that are loaded or
strongly routed at session start.
- Separate automatic context from on-demand references.
- Find bloat and dead routes.
- Look for duplicated rules, stale model-specific instructions, wrappers that
invoke missing skills or commands, broad "read everything" guidance, and large reference files that should be loaded only on demand.
- Classify each issue.
Use the action taxonomy in references/context-surface-taxonomy.md.
- Report findings before edits.
- Prioritize issues by token impact and failure risk.
- Name the canonical owner for every duplicated rule.
- Call out any command wrapper that points to a missing target.
- Apply changes only if asked.
- Keep edits narrow.
- Preserve repo-specific authority boundaries.
- Validate any edited skill frontmatter or command wrapper if a validator is
available.
Output
Use the output template in references/context-surface-taxonomy.md.
If no meaningful bloat is found, say that directly and list the remaining context surfaces that still deserve periodic review.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: mattdweigand-sketch
- Source: mattdweigand-sketch/agent-skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.