incident-retrospective

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs the agent to read an API key from runtime services and embed it in a curl request header, which requires the LLM to handle/output the secret verbatim.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The SKILL.md Step 1 “Verify MCP access” instructs the agent to use the Slack MCP to list recent messages in an incident channel and the Linear MCP to list recent issues, meaning outsider-authored Slack/Linear free text is ingested at runtime via those MCP reads.