The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through the ingestion of external FCS file metadata.
Ingestion points: Untrusted data enters the agent context via the FlowData class when reading FCS files as shown in SKILL.md.
Boundary markers: No delimiters or instructions are provided to the agent to treat extracted metadata as non-instructional data.
Capability inventory: The skill includes file writing and creation capabilities via create_fcs() and write_fcs() as documented in SKILL.md and references/api_reference.md.
Sanitization: No sanitization or validation of the FCS TEXT segment keywords is mentioned before the data is processed or re-exported.

flowio