The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill establishes a protocol for processing handoff payloads, which are untrusted data inputs from other workflows.
Ingestion points: handoff-payload JSON files read from session paths.
Boundary markers: Mandatory JSON Schema (Draft 2020-12) validation is specified to delimit data.
Capability inventory: Scripts for validation using python3, wc, and awk; file system access for logging handoff events and writing JSON files.
Sanitization: Instructions recommend the use of yaml.safe_load and jsonschema.validate to ensure data integrity and prevent exploit payloads.
[Command Execution] (SAFE): The skill provides benign Python and shell one-liners for administrative tasks like UUID generation, token estimation, and schema validation. These commands use standard libraries and safe loading practices (e.g., yaml.safe_load).

workflow-coordinator