The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data such as replication lag metrics, alert statuses, and smoke test results to generate drill reports and make automated go/no-go decisions.
Ingestion points: Monitoring data and verification artifacts (SKILL.md Stage 3, Stage 5; references/restore-time.md Step 7).
Boundary markers: The skill does not define specific delimiters or instructions to treat external monitoring data as untrusted.
Capability inventory: The agent is instructed to use system commands including psql, aws CLI, pg_dump, and pg_ctl.
Sanitization: There is no evidence of sanitization or strict schema validation for the external metrics or status reports ingested.
[COMMAND_EXECUTION]: The skill instructions and reference files contain multiple examples of shell command execution for database management and cloud resource interaction.
Evidence: Extensive use of psql, pg_dump, aws s3, gpg, and pg_restore within SKILL.md and reference files to manage backups and restoration.
[DYNAMIC_EXECUTION]: Technical references include Python code snippets that construct and execute shell commands at runtime using external parameters.
Evidence: references/restore-time.md uses subprocess.run to execute formatted strings for database restoration tasks (e.g., gunzip -c {backup} | psql).

dr-drill-runner