The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from writing samples provided by the user and existing workspace files to generate rules for writing guides.
Ingestion points: The skill reads user-supplied representative samples and external files like ../../references/context-contract.md and TASTE.md to define voice and style rules.
Boundary markers: There are no explicit instructions to delimit user samples or to ignore embedded instructions within that content during the analysis phase.
Capability inventory: The skill has the capability to write to VOICE.md and STYLE.md and can trigger the cw-setup-project skill.
Sanitization: No mechanisms for sanitizing or escaping the content of user-provided samples are specified before they are processed and incorporated into the profile guides.

cw-onboarding