The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection during codebase analysis.
Ingestion points: The agent has full codebase read access via read_file, grep, and codebase_search (specified in SKILL.md and agents/openai.yaml).
Boundary markers: The system prompt lacks delimiters or warnings to ignore instructions embedded in the analyzed data.
Capability inventory: Authorized tools include bash and python execution.
Sanitization: Content from the codebase is not validated or filtered.
[COMMAND_EXECUTION]: The skill utilizes shell-based tools for performance benchmarking.
Authorized execution of bash, python, and the hyperfine utility is documented for triage and analysis tasks.
[EXTERNAL_DOWNLOADS]: The skill accesses official Rust documentation.
Connects to doc.rust-lang.org to retrieve performance-related guidelines, which is a well-known documentation service.

performance-scout