writing-standards
Warn
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: MEDIUMPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a bootstrap script (
bootstrap.py) that the user is instructed to run. This script performs file system operations and modifies the agent's local configuration file.- [PERSISTENCE]: Thebootstrap.pyscript modifies the agent's configuration (settings.json) to installSessionStartandUserPromptSubmithooks. These hooks execute a specific Python script (inject-writing-standards.py) at the beginning of every session and before every user prompt, ensuring that the skill's instructions remain persistently active in the agent's context.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external documents, code, and discussion transcripts, which creates a potential attack surface for instructions embedded in that data. - Ingestion points: User-provided documents, source code, and transcripts intended for review or documentation (specified in SKILL.md Step 2).
- Boundary markers: The skill contains explicit rules (SD-001 through SD-005) directing the agent to abstract the source and avoid direct echoes of its content, establishing a conceptual boundary.
- Capability inventory: The skill reads local files and executes internal Python scripts (bootstrap and hooks).
- Sanitization: No programmatic sanitization is applied; security depends on the agent following the provided abstraction instructions.
Audit Metadata