The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection surface via ingestion of untrusted local data.
Ingestion points: The skill reads project context from the projects/ and docs/strategy/ directories using Read, Glob, and Grep tools in Step 1.
Boundary markers: There are no boundary markers or instructions used to delimit the external data when it is processed or used in subsequent steps.
Capability inventory: The skill has the ability to perform WebSearch, execute sub-agent Task calls, and Write files to the filesystem.
Sanitization: The skill lacks sanitization or validation logic for the content read from files before it is interpolated into prompts for research agents or search queries.

client-discovery