The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted structured data which is used to generate strategy proposals that influence downstream agent decisions. Maliciously crafted input in text fields could potentially manipulate the behavior of agents in the Edge pipeline.
Ingestion points: Strategy iteration history and draft data are ingested via scripts/detect_stagnation.py and scripts/generate_pivots.py.
Boundary markers: The skill does not implement delimiters or 'ignore' instructions to separate processed data from command context.
Capability inventory: The associated scripts perform file system read and write operations but do not involve network or subprocess execution.
Sanitization: Basic sanitization is applied to identifiers through the sanitize_identifier function, but free-text fields like 'thesis' and 'changes' are not escaped or validated.

strategy-pivot-designer