The Agent Skills Directory

[SAFE]: The skill defines a prompt template for formatting code summaries and does not exhibit any malicious behaviors such as obfuscation, persistence, or privilege escalation.
[PROMPT_INJECTION]: The skill processes user-provided code and diffs, which represents an indirect prompt injection surface. However, the risk is negligible as the skill lacks access to tools or execution environments. \n
Ingestion points: Processes 'user-scoped code' and 'unstaged diff' as specified in SKILL.md and agents/openai.yaml. \n
Boundary markers: Absent; the instructions do not provide specific markers to isolate the analyzed code from the agent instructions. \n
Capability inventory: No capabilities are enabled; the skill explicitly sets disable-model-invocation: true in the frontmatter and lists no allowed-tools. \n
Sanitization: None; the skill does not perform validation or escaping of the code content.
[DATA_EXFILTRATION]: The instructions include a security guardrail ('Do not include... secrets') to prevent the accidental display of credentials found within the code being explained.

explain-code