Security audit
Security checks for vulnerabilities and agentic risk
This learning skill needs review because it directs the agent to run unfamiliar code without defining execution safeguards.
For unfamiliar repositories, inspect the code first and require approval before running examples. Use a disposable sandbox without credentials and with restricted filesystem and network access. Choose a new output filename because the scaffold overwrites existing files.
Without declared permissions the skill's intent is opaque and cannot be validated.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
### Stage 0 — Goal inference (reflect, don't ask)
Do NOT ask the user. Instead, ask yourself:
- What is the most common learner intent for this topic? (hands-on use /
technology comparison / secondary development / deep dive / teaching others)
- Who is the most likely reader? (yourself, a colleague, a new onboarder)
The workflow explicitly instructs the agent to run the subject's minimal use case and step through it without requiring a trust assessment or execution isolation. Because the skill covers unfamiliar repositories and reverse engineering, attacker-controlled subject code could execute with the agent's available privileges, potentially accessing credentials, modifying files, or making unauthorized network requests.