The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/collect_issue_context.py uses subprocess.run to execute the sentry CLI. The script passes arguments as a list and does not use shell=True, which follows security best practices to prevent command injection.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process production telemetry (logs, event payloads) that may contain instructions from untrusted sources (e.g., error messages triggered by users).
Ingestion points: Data is ingested through sentry CLI commands such as issue view, issue events, and trace logs, which are then processed by the agent or the collection script.
Boundary markers: The file references/privacy-security.md provides explicit instructions for the agent to treat Sentry data as untrusted evidence and to ignore any instructions embedded within event text.
Capability inventory: The skill has the capability to execute the sentry CLI via subprocess.run and provides instructions for patching the local repository.
Sanitization: The scripts/collect_issue_context.py script implements a redaction function using regular expressions to filter out sensitive patterns such as authorization headers, cookies, tokens, emails, and DSNs before the data is presented to the agent.

sentry-cli-fix-issues