The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates and instructions for creating shell scripts (e.g., zlaunch.sh) that manage complex autonomous vehicle mission lifecycles. This includes spawning multiple background processes, managing rolling parallel execution, and orchestrating temporary file system structures.
[INDIRECT_PROMPT_INJECTION]: The harness templates ingest and parse output files (results.txt) generated by external mission processes to determine test results.
Ingestion points: The results.txt file located in per-case mission work directories (referenced in SKILL.md and references/example-harness-zlaunch.md).
Boundary markers: Scripts utilize specific field parsing (e.g., grade=pass|fail) and shell utility filters like awk and sed to isolate the intended status data.
Capability inventory: The generated harness scripts have the capability to execute shell commands, manage background jobs, signal processes (kill), and perform file system cleanup (rm -rf).
Sanitization: Data processing is restricted to extracting specific result columns, which limits the potential for the ingested data to influence the overall execution flow beyond the test verdict.
[SAFE]: The skill demonstrates safe coding practices, such as promoting scoped process teardown (via the moos_scoped_teardown.sh asset) to prevent killing unrelated system processes, and implementing strict port isolation strategies to avoid network collisions during parallel runs.

moos-ivp-harness-builder