The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation references the use of npx @claude-flow/cli@latest to manage audit records. This command downloads and executes the utility from the npm registry.
[COMMAND_EXECUTION]: The skill orchestrates the execution of multiple local CLI tools via harness sub-commands (oia-manifest, threat-model, mcp-scan) to perform security scans on specified paths.
[DATA_EXFILTRATION]: Audit findings are persisted to a shared memory namespace (metaharness-audit). This is the primary intended behavior of the skill to enable drift detection across multiple runs.
[PROMPT_INJECTION]: The skill processes untrusted local files during the audit phase, presenting an indirect prompt injection surface.
Ingestion points: Local files at the provided <path> are read by the harness scanner tools.
Boundary markers: No explicit delimiters are specified in the markdown instructions.
Capability inventory: The skill uses Bash to execute commands and performs network/storage operations to save audit records.
Sanitization: Sanitization is assumed to be handled by the underlying harness binaries or the oia-audit.mjs script.

harness-oia-audit