The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from external monitoring and incident management platforms, creating a potential surface for indirect prompt injection during report synthesis.
Ingestion points: Incident data is ingested from temporary files such as .pagerduty-oncall-tmp/report.md, .datadog-analyser-tmp/report.md, and .rollbar-reader-tmp/report.md in Step 2.
Boundary markers: There are no explicit delimiters or 'ignore embedded instructions' warnings used when processing the ingested report data to prevent malicious content from influencing the agent.
Capability inventory: The skill has access to the Bash tool for repository analysis and the Skill tool to invoke sub-agents.
Sanitization: No sanitization or validation of the external report content is performed before it is passed to the synthesis step.
[COMMAND_EXECUTION]: The skill uses the Bash tool for repository history analysis and file system maintenance.
Evidence: In steps/step3a-codebase-analysis.md, the skill executes git fetch, git log, and git show to correlate code changes with incident timelines.
Evidence: In steps/step4-present-results.md, the skill uses rm -rf to clean up temporary discovery report folders.
[EXTERNAL_DOWNLOADS]: The skill orchestrates data retrieval from third-party monitoring services via specialized sub-skills.
Evidence: Step 2 involves calling pagerduty-oncall, datadog-analyser, cloudflare-traffic-investigator, and rollbar-reader to gather data from well-known infrastructure and error-tracking services.

pir