The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from various local files to determine the workflow status and suggest the next agent action, creating a surface for indirect instructions.
Ingestion points: scripts/bianpai.mjs reads multiple files including episode-config.json, 00-选题.json, and 02-创作工作稿/创作决策.json.
Boundary markers: Absent in the output provided to the agent.
Capability inventory: The skill can execute shell and Node.js scripts via spawnSync.
Sanitization: The script implements path validation using realpathSync and performs sha256 hash checks and schema validation on ingested files.
[COMMAND_EXECUTION]: The orchestrator script executes several local workflow management scripts (e.g., check-episode-contract.sh, verify-vendor-preflight.mjs) using child_process.spawnSync. The paths for these scripts are derived from the workflow root directory provided by the user.

laohan-bianpai