The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains extensive instructions designed to override the agent's natural output style to ensure it 'passes as entirely human-written'. This includes specific bans on vocabulary (Tiers 1, 2, and 3) and structural patterns typically associated with AI behavior. While intended for quality control, these are pervasive instructions that override base model characteristics.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, posing a risk of indirect instructions influencing the agent.
Ingestion points: The 'Workflow' section in SKILL.md explicitly directs the agent to 'Read the requested draft, source text, or surrounding brief' provided by the user.
Boundary markers: The instructions do not define boundary markers or delimiters to separate the user-provided 'source text' from the skill's instructions.
Capability inventory: The skill has the capability to generate, rewrite, and edit copy, which is then passed back to other skills or the user.
Sanitization: There is no mention of sanitizing or escaping the content of the 'source text' before it is interpolated into the agent's context, potentially allowing instructions embedded in the source text to be executed.

language-standards