show-me-your-work
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies an attack surface by instructing the agent to ingest untrusted data from session transcripts during the audit phase to verify log accuracy.
- Ingestion points: Transcript files located in
agent-transcripts/within the active workspace (referenced inSKILL.md). - Boundary markers: The instructions include explicit guardrails to prevent reading unrelated session data by forbidding global project searches (e.g., 'Don't glob across ~/.cursor/projects/*/').
- Capability inventory: The skill includes file-writing capabilities via
scripts/log.sh. - Sanitization: The logging script proactively sanitizes inputs by prefixing characters that trigger formula execution in spreadsheet software (=, +, -, @) with a single quote.
- [DATA_EXPOSURE]: The skill facilitates the reading of
agent-transcripts/for auditing purposes. While transcripts contain session data, the skill mitigates exposure risks by providing strict scoping instructions that limit the agent's visibility to the current workspace only.
Audit Metadata