skill-publish-log
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a local Python script,
scripts/log.py, to perform all data operations. This includes recording new publish events, querying historical logs, and generating aggregated statistics. The script manages a local JSON file atoutputs/_analytics/publish-log.jsonusing atomic write operations for data integrity.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by storing untrusted user-provided content (such as post titles, tags, and notes) and later re-injecting this content into the agent's context during query and summary operations.\n - Ingestion points: User-supplied input for fields like title, url, tags, and notes via the
recordsubcommand inSKILL.md.\n - Boundary markers: Absent; the LLM is instructed to interpret the JSON output of the script without specific delimiters or instructions to ignore embedded commands within user-generated strings.\n
- Capability inventory: Read and write access to the project's analytics directory; execution of local Python scripts.\n
- Sanitization: No validation or sanitization is performed on user-provided strings to neutralize potential prompt injection instructions before storage or processing.
Audit Metadata