The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by consuming external, untrusted data to drive its debugging and autonomous fix logic. Ingestion points include the structured incident alert JSON inputs and retrieved error log entries. The skill does not employ boundary markers or sanitization to sequester this external content. The capability inventory includes shell execution via bun and npx, file system modifications, and automated git workflow management, which allows the agent to act upon potentially malicious payloads.\n- [COMMAND_EXECUTION]: The skill uses shell commands like git log, grep, bun run typecheck, and npx tsx to investigate the repository and validate configurations. While appropriate for the skill's purpose, these instructions invoke local system tools and scripts.

deco-incident-debugging