The Agent Skills Directory

[COMMAND_EXECUTION]: Executes git log and git diff to provide context on recent development activity. These are read-only operations restricted to the local repository context.
[CREDENTIALS_UNSAFE]: Accesses .env.local to identify required environment variables. The skill contains explicit instructional guardrails: 'names only, never values' and 'Never print .env values', which mitigates the risk of credential exposure.
[PROMPT_INJECTION]: Vulnerable to indirect prompt injection due to the processing of untrusted project files.
Ingestion points: package.json, README.md, supabase/migrations/*.sql, .env.local, and other repository source files.
Boundary markers: Absent; the skill does not use specific delimiters for untrusted file content.
Capability inventory: File reading capabilities across the repository and execution of git commands.
Sanitization: No explicit sanitization or filtering of file content is defined.

workflow-onboard