paw-ps-discovery
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from multiple local ingestion points, creating a surface for potential indirect prompt injection.
- Ingestion points: The agent loads configuration from
{project-root}/.pawbytes/config/config.yamland{project-root}/.pawbytes/config/config.user.yaml, and reads session memory from{project-root}/.pawbytes/prodig-suites/memory/paw-ps-sidecar/index.mdand daily logs. - Boundary markers: There are no defined boundary markers or explicit instructions for the agent to ignore potentially malicious embedded commands within these files.
- Capability inventory: The skill's capabilities are limited to reading and writing text files within the
.pawbytesproject directory. It lacks high-impact capabilities such as network access, shell command execution, or system privilege escalation. - Sanitization: No explicit sanitization, validation, or filtering of the content retrieved from memory files is documented.
Audit Metadata