The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands as part of its mandatory verification process. Step 6 requires running gbrain routing-eval --json to verify structural integrity and uses mktemp and cp to stage files for evaluation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes large routing files (AGENTS.md and RESOLVER.md) which contain trigger phrases that may be user-authored or attacker-controlled.
Ingestion points: Reads configuration files AGENTS.md and RESOLVER.md as specified in the Step 2 selection logic.
Boundary markers: The instructions do not define explicit delimiters or warnings to ignore embedded instructions within the source routing files.
Capability inventory: The skill uses exec to run evaluation tools and write/edit to modify workspace configuration.
Sanitization: There is no mention of sanitizing or escaping the trigger strings extracted from the source files before they are interpolated into the new functional-area dispatcher entries.
[DYNAMIC_EXECUTION]: The skill mandates a verification step that requires running a JavaScript file (harness.mjs) located in an external repository checkout. This involves executing code that is not contained within the skill's own files, representing dynamic execution of external scripts via the node runtime.

functional-area-resolver