The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution by interpolating a user-provided crate name into cargo commands and a local CI script.
Evidence: cargo test -p <crate> --lib --locked in SKILL.md.
Evidence: bash scripts/ci/check-mutation-floor.sh <crate> in SKILL.md.
Evidence: RUSTDOCFLAGS="-D warnings" cargo doc --no-deps -p <crate> in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process documentation and specification files within the workspace, which creates a surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: The agent is instructed to read docs/adr/adr-003-12-gate-admission.md and docs/crate-specs/<crate>.md to perform the audit.
Boundary markers: None. The skill does not define specific delimiters to separate file content from instructions.
Capability inventory: The skill has access to Bash, Read, and Grep tools, allowing it to execute arbitrary shell commands and read any file in the workspace.
Sanitization: The prompt includes a mitigation directive: "Validate the crate name against the manifest before interpolating it into commands."

gate-check