The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a local shell script, validate-records.sh, to validate record integrity and perform security scans. This script leverages standard POSIX utilities like awk, grep, and sed, as well as standard Git commands to verify commit and tag citations.
[PROMPT_INJECTION]: The skill defines a process for ingesting and storing verbatim human judgment and risk statements. While this introduces a surface for untrusted data processing, the implementation includes specific security controls.
Ingestion points: User-authored prose is captured into structured payload fields like Human-Judgment and Human-Accepted-Risk.
Boundary markers: The grammar enforces a block structure using | prefixes to clearly delimit user-provided text from internal metadata.
Capability inventory: The skill is limited to writing local files in the .skills/decisions/ directory and executing the local validation script.
Sanitization: The validation script includes a scanning mode to detect and reject common secret patterns (e.g., AWS credentials, GitHub tokens, Slack tokens) within user-authored sections before publication.

record-verdict