The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze log data from external services and files. The 'AI-Assisted Log Analysis' section provides a prompt template that encourages the agent to process raw log content. This content is untrusted and could contain hidden instructions designed to override agent behavior during the analysis phase.
Ingestion points: File reads from logs.json and logs/*.json, as well as manual log input via the [paste logs] placeholder in SKILL.md.
Boundary markers: The skill suggests wrapping log content in markdown code blocks (triple backticks), which provides basic delimitation but does not prevent adversarial instructions within the data.
Capability inventory: The skill utilizes shell commands (grep, jq) and TypeScript logic to filter and reconstruct timelines from file data.
Sanitization: There is no evidence of sanitization or filtering to detect or strip natural language instructions embedded within the log messages before processing.
[COMMAND_EXECUTION]: The skill includes several examples of Bash commands using grep and jq to perform file system operations and data manipulation on log files. While these are standard diagnostic tools, they involve direct shell execution on local data.

log-forensics