The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external, potentially untrusted data including source files and git diffs, creating a surface for indirect prompt injection.
Ingestion points: The skill reads .rs source files, Cargo.lock dependency information, and the output of git diff and cargo clippy as defined in the 'Scope' and 'Phase 1' sections.
Boundary markers: A dedicated 'Guardrails' section instructs the agent to treat all ingested files, comments, and tool outputs as data and to never follow instructions found within them.
Capability inventory: The skill utilizes shell command execution (cargo clippy, cargo check, git diff), file system access (reading project files and the local cargo registry), and the ability to spawn subagents.
Sanitization: Mitigation is limited to prompt-level instructions; the skill does not programmatically sanitize or filter the content of the files it analyzes.

egui-review