@majiayu000/tool-use-guardian
BFREE — Intelligent tool-call reliability wrapper. Monitors, retries, fixes, and learns from tool failures. Auto-recovers from truncated JSON, timeouts, rate limits, and mid-chain failures.
Install
agr install @majiayu000/tool-use-guardian --target claudeWrites 2 files into .claude/skills/, pinned to git-5319ca3c.
- .claude/skills/tool-use-guardian/SKILL.md
- .claude/skills/tool-use-guardian/metadata.json
Document
name: tool-use-guardian description: "FREE — Intelligent tool-call reliability wrapper. Monitors, retries, fixes, and learns from tool failures. Auto-recovers from truncated JSON, timeouts, rate limits, and mid-chain failures." category: reliability risk: safe source: community date_added: "2026-03-13" author: christopherlhammer11-ai tags: [reliability, tool-use, error-handling, retries, recovery, agent-infrastructure] tools: [claude, cursor, codex, gemini, copilot, windsurf, antigravity]
Tool Use Guardian
Overview
The reliability wrapper every AI agent needs. Monitors tool calls, auto-retries failures, fixes truncated responses, and learns which tools are unreliable — so you never lose your chain of thought.
Free forever. Built by the Genesis Agent Marketplace.
Install
npx skills add christopherlhammer11-ai/tool-use-guardian
When to Use This Skill
- Use when tool calls return truncated or malformed JSON
- Use when APIs timeout or rate-limit your agent mid-task
- Use when a multi-step chain breaks partway through
- Use when you need automatic retry logic without writing it yourself
- Use for any agent workflow that depends on external tool reliability
How It Works
Step 1: Pre-Call Validation
Before every tool call, Guardian validates:
- Required parameters are present and correctly typed
- The tool is not marked as "unreliable" from previous failures
- Request size is within known limits
Step 2: Failure Classification
When a tool call fails, Guardian classifies the failure into one of 9 categories:
| Failure Type | Recovery Action |
|---|---|
| Truncated JSON | Re-fetch with pagination or smaller chunks |
| API Timeout | Retry once with simpler request, then decompose |
| Rate Limit (429) | Exponential backoff, max 3 retries |
| Auth Expired | Flag for user intervention |
| Mid-chain Break | Resume from last successful checkpoint |
| Error-as-200 | Detect {"error": "..."} disguised as success |
| Schema Mismatch | Attempt auto-coercion, warn if lossy |
| Network Failure | Retry with jitter, max 2 attempts |
| Unknown Error | Log full context, escalate to user |
Step 3: Chain Protection
For multi-step tool chains, Guardian maintains checkpoints. If step 4 of 7 fails, it resumes from step 4 — never restarts from scratch.
Step 4: Learning
Guardian tracks failure patterns per tool. After 3+ failures of the same type, it marks the tool as unreliable and suggests alternatives.
Best Practices
- ✅ Let Guardian wrap all external tool calls automatically
- ✅ Review Guardian's reliability reports to identify flaky tools
- ✅ Use checkpoint recovery for long chains
- ❌ Don't disable retry logic for rate-limited APIs
- ❌ Don't ignore repeated failure warnings
Related Skills
@recallmax- Long-context memory enhancement (also free from Genesis Marketplace)
Links
- Repo: https://github.com/christopherlhammer11-ai/tool-use-guardian
- Marketplace: https://genesis-node-api.vercel.app
- Browse skills: https://genesis-marketplace.vercel.app
Limitations
- Use this skill only when the task clearly matches the scope described above.
- Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
- Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.
Repository README
Describes majiayu000/claude-skill-registry-data as a whole, which may contain artifacts other than this one. Where this artifact had no useful description of its own, its summary was taken from here.
Claude Skill Registry (Data)
This repo contains the archived skill contents (the heavy, browsable skill files).
Canonical layout
- Category folders at repo root (e.g.
development/,documents/,data/, ...) - Each skill lives under a category:
<category>/<skill>/SKILL.md+<category>/<skill>/metadata.json - Case conflicts are resolved with
{name}-{owner}-{repo}suffixes (fallback:-{short-hash}).
Archive status
- Live badges above are sourced from
claude-skill-registry-corestats.json. - Counts in this README are intentionally dynamic, not hardcoded.
- If the badges look stale, refresh the
corebuild/index pipeline rather than editing numbers here.
Where the index + site live
- Core repo: https://github.com/majiayu000/claude-skill-registry-core
- Main repo (merged publish artifact): https://github.com/majiayu000/claude-skill-registry
Trustgrade B
- passBody integrity
Whether the stored document is plausibly the kind of file the artifact declares, rather than something fetched by mistake.
- passType matchnot applicable to this artifact type
Whether the artifact is really the kind of thing its metadata claims it is.
- passFreshness
How long since the source repository was last pushed to.
- passPrompt injection
Scans the artifact's own text for instructions aimed at your agent rather than at you.
- warnLicenseno SPDX license detected
Whether the source repository declares an SPDX license permissive enough to redistribute.
How the grade is calculated
Each check contributes 0 points when it passes, 1 when it warns, and 2 when it fails. The total maps to a letter:
- Aevery check passed
- Bone warning
- Ctwo warnings
- Dprompt injection or body integrity failed, or three warnings
- Fone of those failed, and something else is wrong
These are automated hygiene checks, not a security audit, and not a dependency or vulnerability scan. A grade of A means nothing was flagged — not that the artifact is safe.
Versions
git-5319ca3c8bbb2026-07-31