The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection, as it is configured to ingest external data and has tools that could be manipulated by that data.
Ingestion points: The agent can ingest untrusted data from the web using the WebFetch tool defined in the allowed-tools section of SKILL.md.
Boundary markers: Absent. The instructions do not include specific delimiters or warnings to the agent to ignore instructions embedded in fetched external content.
Capability inventory: The skill allows the use of Bash(gh api *) and Bash(gh search *) in SKILL.md, which enables the agent to interact with GitHub repositories.
Sanitization: Absent. There is no instruction to sanitize or validate external content before processing it.
[SAFE]: The instructional content regarding the Algemene Verordening Gegevensbescherming (AVG) is accurate and focuses on privacy best practices like pseudonymization, data minimization, and lawful processing.
[SAFE]: Code examples for BSN (Burgerservicenummer) handling follow security standards, advising against logging sensitive data and recommending encryption for storage.
[SAFE]: No obfuscation, hardcoded credentials, or persistence mechanisms were detected in the skill files.

avg-privacy