The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized network operations were detected in the skill's scripts or documentation.
[DATA_EXPOSURE]: The skill manages its own configuration by reading and writing to standard user directories (~/.config/research-ops/product-research.json). This is a legitimate implementation of persistent settings for the tool's operation and does not involve accessing sensitive system credentials or private data.
[INDIRECT_PROMPT_INJECTION]: The insight_synthesizer.py script and the ar_evaluator.py tool ingest research observations from an external JSON file (observations.json).
Ingestion points: Untrusted data enters the agent context through the observations array in the input file.
Boundary markers: None are explicitly used in the scripts.
Capability inventory: The skill has the capability to write configuration files to the local file system via config_loader.py and onboard.py.
Sanitization: None. While the script logic is deterministic and safe, if the agent interprets the processed observations as instructions, it could be a surface for indirect prompt injection.
[METADATA_POISONING]: A discrepancy exists between the author name in the skill's metadata (claude-code-skills) and the provided author context (alirezarezvani). While potentially misleading about the skill's origin, it does not represent a functional security threat.

product-research