The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's initialization process relies on reading various existing files within a repository to infer configuration and project state. This ingestion of untrusted data represents an indirect prompt injection surface.
Ingestion points: In SKILL.md, the "Explore" phase instructs the agent to read project files including CONTEXT.md, CONTEXT-MAP.md, CLAUDE.md, AGENTS.md, and Architecture Decision Records (ADRs) found in docs/adr/.
Boundary markers: The skill does not define specific delimiters or instructions to the agent to ignore potentially malicious embedded instructions within these repository files.
Capability inventory: The skill performs file writes to create or update docs/agents/setup-state.json, issue tracker documentation, triage label definitions, and .cursor/rules/fancy-skills.mdc. It also modifies existing CLAUDE.md or AGENTS.md files.
Sanitization: No validation or sanitization logic is specified for the content read from the repository's documentation files.
[EXTERNAL_DOWNLOADS]: The skill templates and instructions reference the use of external version control interfaces and their respective command-line tools.
Evidence: The templates issue-tracker-github.md and issue-tracker-gitlab.md instruct the agent to use the GitHub (gh) and GitLab (glab) CLI tools for issue management. The GitLab template specifically includes a reference link to the official GitLab CLI repository at https://gitlab.com/gitlab-org/cli. These references target well-known and established development services.

fancy-init