tsg-flow-next
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from various local project files (e.g.,
flow-state.json,capabilities.md, and files listed in theartifactssection) to guide the agent's next actions. This data ingestion represents a potential attack surface if these files contain malicious instructions. - Ingestion points:
flow-state.json,backlog/capabilities.md, and markdown documents referenced in the state file (e.g., domain maps, PRDs). - Boundary markers: The skill relies on structured metadata parsing via a helper script, but does not implement explicit delimiters or instruction-ignore warnings for the natural language content it reads.
- Capability inventory: The skill executes a local Python validation script and has the authority to transition state in
flow-state.jsonand chain other creation skills. - Sanitization: The
validate_state.pyscript performs structural and versioning checks on metadata but does not sanitize or filter the natural language text for prompt injection patterns. - [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script,
scripts/validate_state.py, to perform mechanical validation of the project's current state and dependencies.
Audit Metadata