The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-provided writing drafts for revision and diagnosis. This functionality creates an attack surface for indirect prompt injection, where malicious instructions could be embedded in the text being audited to influence the agent's behavior. The skill includes specific safeguards like 'voice-guards' and 'narrative-honesty-guards' to preserve the original text's integrity, which helps mitigate accidental overrides.
[COMMAND_EXECUTION]: The skill includes a shell script evals/run-static-checks.sh that executes local repository scripts using python3. This script is used for static validation of the skill's configuration and templates during the development and testing phase. These commands do not process untrusted runtime arguments or external data, and are scoped to the development environment.

writing-audit