The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, creating a surface for indirect prompt injection attacks where malicious instructions hidden in data could influence agent actions.
Ingestion points: The agent ingests user requests (SKILL.md), reproduces bugs from potentially untrusted logs (debugging/systematic/SKILL.md), and evaluates pull request descriptions and comments (quality/code-review/SKILL.md).
Boundary markers: The instructions lack explicit boundary markers or directives for the agent to ignore embedded instructions when reading external codebase files or log outputs.
Capability inventory: The skill leverages powerful behaviors including run_command for system discovery and testing, read_file for codebase analysis, and write_file for implementing code changes.
Sanitization: There is no evidence of input sanitization or validation protocols applied to ingested data before it is used to influence the agent's reasoning or execution flow.
[COMMAND_EXECUTION]: The skill performs systematic discovery of the host system using various command-line tools.
Evidence: During the 'Hard-Gate' scan phase, the skill executes df -h (disk space), free -m (memory usage), ps aux (process monitoring), and git log (version control history) to gather context about the environment.

kilo-kit-core