personal-branding-authority
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources, including web-fetched company pages and user-provided CVs or profile text. It incorporates robust defensive prompts that instruct the agent to treat this data exclusively as content and never as instructions, providing a specific protocol for handling and reporting any detected injection attempts.\n
- Ingestion points: Company 'About' pages, user CVs, and profile text (SKILL.md, Step 1 and 2).\n
- Boundary markers: Explicit instructions to ignore embedded commands and a mandatory reporting line (SKILL.md, Section 4, Step 7).\n
- Capability inventory: Can perform web fetches and execute a local Python script.\n
- Sanitization: Instructions to strictly separate data from instructions.\n- [COMMAND_EXECUTION]: The skill uses a local Python script
scripts/count_fields.pyto count words and characters in drafted assets.\n - Evidence: The script is part of the skill package and is invoked using
python3. It includes input validation for command-line arguments to prevent malformed inputs.\n- [EXTERNAL_DOWNLOADS]: The skill is designed to fetch content from employer 'About' pages to verify company information.\n - Evidence: Instructions in
SKILL.mdspecify fetching official company URLs and quoting them with dates and URLs in a fact sheet.
Audit Metadata