The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted diagnostic logs and service data provided in a bundle. While the instructions include specific rules to filter volatile content and mask credentials, the ingestion of external data presents a surface for indirect instructions embedded in logs to influence agent behavior.\n
Ingestion points: The agent reads file content from bundle/logs/ as defined in SKILL.md and reference.md.\n
Boundary markers: There are no explicit delimiters or specific 'ignore embedded instructions' warnings for the log content being processed.\n
Capability inventory: The skill allows the use of Bash (for grep, find, and gh search), Read (filesystem access), and Write (report generation).\n
Sanitization: The rules/report-evidence-per-finding.md file mandates that the agent identify and mask any credential-shaped strings before including log excerpts in the final report.\n- [COMMAND_EXECUTION]: The skill makes use of the Bash tool to execute grep and find for log analysis and the gh CLI for searching GitHub issues in the opsmill/infrahub repository. These operations are essential for the skill's purpose and are constrained by instructions to read-only diagnostic activities.

infrahub-analyzing-diagnostics