Security audit

Hermes Doctor

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a legitimate Hermes troubleshooting guide, but it needs review because it directs agents to inspect, copy, back up, and modify API keys and auth files without consistently warning users how to protect them.

Install only if you want a troubleshooting skill that may guide an agent through Hermes credential and configuration repair. Before following its steps, confirm the exact HERMES_HOME, prefer read-only diagnostics first, redact API keys and Authorization headers, avoid pasting raw terminal output into chats or tickets, encrypt or tightly restrict any backups, and require explicit approval before editing .env/auth.json/config.yaml, deleting registry values, running --fix commands, killing processes, or moving credentials between machines.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (10)

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The README instructs users to inspect request dump files and Authorization headers during troubleshooting, but it does not warn that these artifacts may contain live bearer tokens or other sensitive authentication material. In a diagnostic skill, normalizing manual inspection of credential-bearing logs increases the chance of accidental disclosure through screenshots, copy/paste, issue reports, or committing dumps to source control.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The instructions tell users to inspect request dump files for Authorization header values without warning that these artifacts may contain live or partially masked secrets. In a troubleshooting skill focused on auth failures, this increases the chance users will expose API keys in terminals, shell history, screenshots, logs, or copied snippets during diagnosis.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The document instructs users to manually place live API keys into .env and auth.json in plaintext and even recommends dual-writing them for reliability, but does not warn about plaintext secret storage risks. This can lead to credential disclosure through backups, file syncing, local compromise, accidental commits, support bundles, or shared workstation access.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The document instructs users to archive and transfer the entire HERMES_HOME, explicitly including sensitive files such as .env and auth.json, but does not place a clear warning immediately alongside those steps about the high sensitivity of the backup artifact. This creates a real risk of credential leakage through insecure storage, accidental sharing, or transfer over poorly controlled channels during migration.

Missing User Warnings

Medium

Confidence: 85% confidence
Finding: The emergency recovery steps rename and replace config/auth files to restore service, but they do not clearly warn users up front that these actions can discard current authentication state, overwrite recovery context, or force regeneration of credentials. In a troubleshooting skill, users are likely to copy-paste commands under pressure, which makes undocumented destructive behavior more dangerous.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The document explicitly instructs users to place live API keys on the command line and append them into `.env`, then back up `.env` and `auth.json`. That creates real secret-exposure risk through shell history, process listings, copied backup files, and manual editing of credential stores, especially because the text does not warn about these operational security consequences.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The document explicitly recommends `hermes doctor --fix`, which makes automatic repairs to configuration and runtime state, but it does not prominently warn users that this can modify files, migrate config, checkpoint WALs, or rebuild symlinks. In a troubleshooting skill, users are primed to copy-paste commands, so omission of a clear change-management warning can lead to unintended local state changes or destructive remediation on the wrong Hermes home/profile.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The document tells users to inspect `request.headers.Authorization` in request dump files, which directly encourages viewing and potentially storing bearer tokens or API keys in plaintext artifacts. In a troubleshooting skill focused on auth/provider issues, this is more dangerous because users are already handling live credentials and may share dumps in tickets, chat, or screenshots without realizing the exposure risk.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The checklist tells users to print environment variables and inspect auth material such as API_KEY-related values and credential pool contents directly to the terminal without warning about secret exposure. In practice this can leak credentials into shell history, terminal scrollback, screen recordings, shared consoles, support screenshots, or logs copied into tickets during troubleshooting.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The checklist instructs users to delete credential-related registry entries directly with `reg delete` but does not warn about disruption, scope, or recovery implications. A user can remove the wrong variable, break working authentication or path resolution, and lock themselves out of the correct Hermes environment, especially on Windows where HKCU environment overrides are easy to misuse.

Static analysis

No suspicious patterns detected.