open-knowledge-pack-software-lifecycle-write-a-postmortem

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a tool named exec to perform local file system operations, specifically ls, cat, and grep. These are used to list prior postmortems and search for keywords within them.
  • [EXTERNAL_DOWNLOADS]: The skill references its official repository at https://github.com/inkeep/open-knowledge for installation and setup. This is consistent with the author's identity and stated purpose.
  • [PROMPT_INJECTION]: The skill includes a "HARD GATE" instruction that directs the agent to override user requests if they contain blame-oriented framing, reframing them into system-focused questions instead. This is a functional policy designed to enforce a "blameless" postmortem culture.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted external data such as chat transcripts, logs, and deployment history, which could contain adversarial instructions.
  • Ingestion points: Untrusted data enters the agent context via host native tools during the evidence gathering phase (Step 1 in SKILL.md).
  • Boundary markers: Absent. The instructions do not specify the use of delimiters or clear separation markers for the gathered external data.
  • Capability inventory: The agent possesses write and edit tools to modify the file system and an exec tool to run shell commands (ls, cat, grep).
  • Sanitization: Absent. There are no explicit instructions to filter, escape, or sanitize the content of logs or chat transcripts before analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 02:16 AM
Security Audit — agent-trust-hub — open-knowledge-pack-software-lifecycle-write-a-postmortem