Audit Hermes Agent Skills - Setup Guide¶

Source: cnife/skills GitHub: github.com/cnife/skills (repo: 25 skills, 354 installs) Category: Skill Management / Hygiene First Seen: April 22, 2026 Security: Gen Agent Trust Hub Pass · Socket Pass · Snyk Pass

A skill-usage auditor for Hermes Agent. It measures how often each installed skill is actually used, identifies long-dormant skills, and produces a cleanup decision report. The skill queries Hermes' internal API (_find_all_skills, _read_manifest, HubLockFile) for authoritative skill source classification (hub/builtin/local/external) and combines it with filesystem scanning to locate real directory paths.

Note: the skill's UI and report are in Chinese (技能审计报告).

Installation¶

npx skills add https://github.com/cnife/skills --skill audit-hermes-agent-skills

The skill requires uv for dependency management - all commands run through uv run.

Usage¶

Step 1 - Generate the audit report¶

uv run ~/.hermes/skills/audit-hermes-agent-skills/scripts/audit.py

Prints a console overview and writes 技能审计报告.xlsx (skill audit report) to the current directory.

Step 2 - Review in Excel¶

The XLSX lists every installed skill with source classification and usage data. Mark keep/remove decisions per row.

Step 3 - Clean up¶

Remove skills marked for deletion, guided by the source classification:

Source	Meaning	Cleanup Path
hub	Installed from a registry	`npx skills remove <repo>`
builtin	Shipped with Hermes	Do not remove - exclude from reports
local	User-created	`rm -rf` the skill directory
external	Manual/other installs	Remove via the original install path

Prerequisites¶

Requirement	Details
Hermes Agent	Installed and accessible from the shell
uv	Python package manager (`curl -LsSf https://astral.sh/uv/install.sh \| sh`)
Excel-compatible viewer	For the XLSX report

CorpusIQ Use Cases¶

Use Case	How
Quarterly skill audits	Automate the skill-pruning cycles CorpusIQ runs manually today
Usage evidence	Decide skill deletions on real usage frequency instead of memory
Source hygiene	Identify external/hub skills that drifted from their install source
Bloat detection	Catch skill sprawl before context and startup costs degrade

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations / Verification¶

Report UI is in Chinese - decisions are still readable as skill names, sources, and counts
Requires Hermes' internal APIs to be accessible; if Hermes relocates them, the script may need updates
Verify: uv run ~/.hermes/skills/audit-hermes-agent-skills/scripts/audit.py exits 0 and produces 技能审计报告.xlsx

Powered by CorpusIQ