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 skills add AmineDjeghri/personal-os-setup --skill hermes-instance-auditgit clone --depth 1 https://github.com/AmineDjeghri/personal-os-setupWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/aminedjeghri/personal-os-setup/hermes-instance-audit)<a href="https://agentmods.dev/skills/aminedjeghri/personal-os-setup/hermes-instance-audit"><img src="https://agentmods.dev/badge/skills/aminedjeghri/personal-os-setup/hermes-instance-audit/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/aminedjeghri/personal-os-setup/hermes-instance-audit"><img src="https://agentmods.dev/badge/skills/aminedjeghri/personal-os-setup/hermes-instance-audit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00018 | $0.05861 |
| Opus 5.5 | $0.00007 | $0.02344 |
| Sonnet 5.5 | $0.00004 | $0.01172 |
| Haiku 4.5 | $0.00002 | $0.00586 |
Grade A, and why
hermes-instance-audit scanned grade A with 1 finding 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 18d ago.
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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- **Prove absence against a CONTROL, never from one probe's silence.** A check that returns the same negative for items known to work is a broken probe, not a finding: a `dig`/`getent` sweep returned nothing for EVERY ho How it starts
The opening of the file, as written. The whole thing — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hermes Instance Audit (read-only)
Answer "what do we have / is anything wrong / how does it compare upstream" about a live Hermes Agent installation without changing it. Read-only unless the user explicitly asks for a change.
When to use
- "Check what we have in terms of skills, plugins and more" · "is anything wrong with my Hermes"
- "Compare it with the latest version / release" · "did they remove the default installed stuff"
- "Where did skill X go / why is it disabled / can I get deleted skills back"
- "Why does the dashboard say X" · "why is context at X% / do these tools eat too much"
- A pasted add-on/gateway BOOT LOG — "why is /dashboard/ dead", "what does this error mean" → boot-log triage below, depth in
references/addon-boot-log-triage.md - Not for installing, vendoring, merging or wiring skills →
agent-skills-architecture.
Hard rules
- Read-only:
read_file,search_files,web_extract. Never editconfig.yaml(onlyhermes config set), never run a sync by hand, never restore a skill without an explicit go-ahead. The deliverable of an audit is the EXPLANATION — what exists, what it does, what (if anything) needs doing — and "nothing needs doing" is a valid, expected answer. Do not close with install/enable/cleanup pitches or a menu of optional upgrades: he reads an unrequested proposal as a defect in the answer, and only wants options when he asks for them. - Approval gate: with
approvals.mode: manual,terminal/execute_codecalls can sit at the consent gate and time out (BLOCKED). An audit is doable on file tools + read-only plain-shell one-liners: singlemkdir/cp/tar/find/diff/md5sum/greplines pass, whileexecute_code,python -c, heredocs and multi-line loops/scripts time out even when read-only (verified 2026-08). Keep shell work to one-liners; do the rest with file tools. Never retry a blocked call — report it and let the user say "prompt me again"; when he does, re-issue the SAME call, which then runs (the gate was the problem, not the call). Never reach the same outcome through a substitute tool in the meantime. .envis access-denied toread_file(credential store) — audit keys from memory/session context instead.- web_extract caches of big GitHub API JSON are ONE giant line:
read_filecannot page them (nonext_offset). Usesearch_fileswith context, or fetch a single release/tag endpoint (~60KB, fits one read).
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/addon-boot-log-triage.md 7.4 KB
- references/platform-adapter-dependencies.md 4.7 KB
- references/plugins-platforms-channels.md 9.5 KB
- references/release-history.md 6.1 KB
- references/skill-loading-resolution.md 9.5 KB
- references/skills-sync-and-bundling.md 4.0 KB
- scripts/check_bundled_manifest.py 4.2 KB runs code
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.
- 18d ago Changed · -1 lines 6987dd990800
- 27d ago First seen · 145 lines · 18 tokens per session scan A 75f2e120ab4c
hermes-instance-audit is a skill published in the GitHub repository AmineDjeghri/personal-os-setup (654 stars, last pushed today), licensed MIT. It adds 18 tokens to every session and 5,861 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-19.
Other skills, from other repositories
hula-skill
HuLa project skill for frontend (Vue 3 + Vite + UnoCSS + Naive UI/Vant), backend (Tauri v2 + Rust + SeaORM/SQLite), full-stack flows, and build/release work. Use when the user mentions hula or HuLa or requests changes in this repository; after triggering, ask which scope (frontend/backend/fullstack/build-release) to…
flow-launcher
Build Flow Launcher plugins with C# or Python JSON-RPC, configure plugin manifests, and integrate Everything search.
microsoft-powertoys
Configure PowerToys FancyZones, Run, Keyboard Manager, and Awake; automate settings and troubleshoot Windows utility behavior.
eartrumpet
Inspect and route per-application audio with EarTrumpet, WASAPI, and pycaw; troubleshoot audio sessions and output devices.
wsl-bash-crlf-or-tempfile
When a Windows shell (PowerShell/cmd) feeds a bash script into WSL, CRLF line endings can corrupt the first shell builtin; force LF or pipe via a temp file.
wsl-networking-mode-dns-fallback
When WSL's mirrored networking fails and falls back to "None", plus /etc/wsl.conf has generateResolvConf=false, the distro has no DNS; fix both layers.