pipa-daily-shutdown
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill architecture processes data from external sources such as project trackers and calendars, creating a potential surface for indirect prompt injection. This risk is mitigated by explicit instructions in 'references/gotchas.md' to treat retrieved records as untrusted and to ignore any embedded instructions. This security behavior is actively tested in the skill's evaluation suite.
- Ingestion points: Data retrieved from external trackers, calendars, and code hosting platforms via tools.
- Boundary markers: Instructions to ignore instructions in retrieved records are present in the 'gotchas.md' reference file.
- Capability inventory: Reading app-specific configuration files in '~/.pipa/' and proposing external writes to messaging or task platforms.
- Sanitization: The agent is instructed to treat all external data as read-only records and to verify all proposed writes with the user.
- [DATA_EXFILTRATION]: The skill reads local files in the '~/.pipa/' directory to understand user goals and available tool connectors. These files are specific to the skill's operational context and do not include sensitive system credentials. All external writes require a separate approval step from the user, preventing unauthorized data transmission.
Audit Metadata