analyze-team-session
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches latest best practices and token cost information from the official documentation at
code.claude.com. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided markdown session exports which may contain untrusted or adversarial content intended to influence the agent's behavior.
- Ingestion points:
SKILL.md(Step 1) instructions the agent to read the full markdown session export file provided in the argument. - Boundary markers: The workflow does not include explicit delimiters or instructions for the agent to ignore potentially malicious embedded prompts within the session transcript.
- Capability inventory: The skill possesses file-read and file-write access (specifically to
.claude/output/), as well as network capabilities viaWebSearchandWebFetchtools. - Sanitization: No sanitization, validation, or filtering of the transcript content is performed before the agent extracts and evaluates the data.
Audit Metadata