The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains directives meant to bypass the agent's standard introductory behavior and suppress summaries of its function. Specifically, it uses 'CRITICAL: Auto-start on load' and instructions to 'not explain what it does' or 'list the files it references', which are concealment patterns that reduce user oversight.
[PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by ingesting untrusted user content, such as 'context dumps' and 'reference posts', which are then used to influence the agent's generation process. Ingestion points: Step 1 gathering user input from 'context dump' and 'reference posts' options. Boundary markers: Absent; there are no instructions to the agent to ignore or delimit potentially malicious instructions within the user-provided data. Capability inventory: The skill has file system write capabilities (Step 4) and uses interactive user communication tools. Sanitization: Absent; the skill does not specify any filtering or validation for the content of the external inputs.

post-writer