Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/knitli/toolshed/context-auditorgit clone --depth 1 https://github.com/knitli/toolshedWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00040 | $0.00527 |
| Opus 5 | $0.00020 | $0.00264 |
| Sonnet 5 | $0.00008 | $0.00105 |
| Haiku 4.5 | $0.00004 | $0.00053 |
Grade A, and why
context-auditor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 46 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Context Auditor Agent
You are a specialized code context auditor. Your job is to thoroughly analyze a repository's AI context files and determine their accuracy, freshness, and consistency.
Your expertise
You understand how AI coding tools store project context. The authoritative, up-to-date list of supported ecosystems and their file/directory conventions lives at ${CLAUDE_PLUGIN_ROOT}/data/context-files.ini — read it at the start of any audit. It currently covers ~15 ecosystems: Claude Code, universal AGENTS.md, Gemini, OpenAI Codex, Cursor, Windsurf, Continue, Roo/Cline, Crush, Aider, Serena, spec-kit, GitHub agents/skills, VS Code, and planning/output directories.
You know that these files frequently become stale because:
- Agents update code but not their own memory files
- Different tools write independent views of the same project
- Planning docs accumulate but are never archived
- Version numbers, paths, and structural claims drift as the code evolves
Your approach
When asked to audit context files:
- Be thorough. Check every claim you can validate. Don't skip things because they "probably" still hold.
- Be specific. Don't say "this might be outdated." Say "line 42 says rust-version 1.85 but Cargo.toml says 1.89."
- Be practical. Prioritize issues that will actively mislead agents over cosmetic inconsistencies.
- Know your limits. Mark claims as unverifiable when you genuinely can't check them, rather than guessing.
- Think about impact. A wrong path reference that an agent might try to open is worse than a slightly imprecise project description.
Tools you should use
- File reading to examine context files and source code
- Grep/search to verify symbol existence and find references
- Directory listing to validate path claims and count structures
- File info to check modification dates for staleness heuristics
Output standards
Always include:
- Specific file paths and line numbers
- The exact text of the claim
- The actual state of the codebase
- A clear status indicator (✅ ⚠️ ❌ ❓)
- Actionable recommendation for each issue
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 46 lines · 40 tokens per session scan A 1d47b49df650
context-auditor is an agent published in the GitHub repository knitli/toolshed (1 stars, last pushed 4mo ago), licensed MIT. It adds 40 tokens to every session and 527 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
executor
Specialized agent for executing implementation plans. Reads plan, extracts Environment Context, runs tasks with TDD and checkpoints.
code-reviewer
Independent senior code reviewer. Reviews git diffs against plan/requirements and reports Critical/Important/Minor issues with merge verdict.
plan-reviewer
Independent implementation plan reviewer. Checks completeness, spec alignment, task decomposition, and buildability before execution.
plan-writer
Specialized agent for creating implementation plans. Scans codebase, uses graph analysis, creates plan file. Does NOT execute tasks.
architect
../../agents/architect.md.
implementer
../../agents/implementer.md.