sub-agent-verification

Verify sub-agent output with syntax checks, smoke tests, and git review before reporting completion.

1|Updated Mar 15, 2025
One-click install
npx skills add https://github.com/adikpb/dotfiles --skill sub-agent-verification-adikpb
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sub-agent-verification
Source: https://github.com/adikpb/dotfiles/tree/main/.hermes/skills/autonomous-ai-agents/sub-agent-verification
Command: npx skills add https://github.com/adikpb/dotfiles --skill sub-agent-verification-adikpb

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Sub-agents and delegated coding tools like opencode often report success while their actual output contains syntax errors, broken escaping, missing imports, or files written to wrong paths. This Skill enforces independent verification of delegated work before claiming a task is done. ## Core Features & Use Cases - Layered Verification Pipeline: Run syntax/compilation checks, module import tests, functional smoke tests, project test suites, git diff review, and content marker greps against sub-agent output. - Truncated Report Recovery: Recover full sub-agent reports from delegation cache files when batch results are truncated, using the included extraction script. - Convergence Audit Loops: Run repeated recon-verify-fix rounds with blind final rounds and live end-to-end gates until sub-agents report zero new findings. - Use Case: After delegating a multi-file feature build to opencode, run the import test, smoke script, and git diff review to catch a foreign-key schema error and a placeholder npm test script before telling the user the work is complete. ## Quick Start Verify the sub-agent's changes by running the syntax check, import test, and git diff review before confirming the task is done.

Frequently Asked Questions about sub-agent-verification

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify sub-agent output before marking a task done?▼

Run layered checks yourself: syntax parsing, module import in the project venv, a functional smoke test, the project's test suite, git diff review, and content marker greps. Never rely on the sub-agent's self-reported success as evidence.

How to recover truncated delegate_task batch reports?▼

Full reports live on disk in ~/.hermes/cache/delegation/subagent-summary-*.txt files. Use the included extract_delegation_report.py script to pull the JSON-escaped report key, or check /private/tmp for agent-written report files.

Should I fix sub-agent errors myself or re-delegate the task?▼

Fix syntax errors and small adjustments like styling or naming yourself since they are fast mechanical fixes. Re-delegate with a clearer spec only when the sub-agent took an entirely wrong approach to the problem.

Why does opencode output pass syntax checks but still break?▼

Syntax and import checks miss issues like broken JavaScript template literal escaping inside Python strings or missing DOM markers. Content verification, such as grepping live page output for expected markers, catches these failures.

When should a recon-verify-fix audit loop stop?▼

Stop when a round returns no new findings from all agents, since convergence is per-agent rather than global. A final blind round without prior-round briefing catches regressions that briefed rounds cannot detect.