The Agent Skills Directory

[Indirect Prompt Injection] (MEDIUM): The skill exhibits a significant attack surface for indirect prompt injection. 1. Ingestion points: Untrusted log files processed by the scripts/preprocess.py script. 2. Boundary markers: Absent; there are no specified delimiters or instructions to ignore embedded commands within the log content. 3. Capability inventory: The skill generates 'Smart Insights' and 'Recommendations' which are subsequently interpreted by the agent to provide root cause analysis and suggestions. 4. Sanitization: No automated sanitization is described; the skill relies on manual user caution to avoid sensitive data exposure. Maliciously crafted log entries could include hidden instructions that manipulate the agent's reasoning or induce it to provide harmful recommendations.
[Data Exposure] (MEDIUM): The skill is explicitly designed to identify and extract 'ak' (Access Keys) and session identifiers. While intended for debugging and auditing, this behavior centralizes highly sensitive credentials into plaintext structured files (entities.md, analysis.json), increasing the risk of credential exposure if the logs contain active secrets.

log-analyzer