The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as HTTP request metadata (headers, URLs, methods) and application logs for monitoring purposes.
Ingestion points: TailEvent data structures in templates/tail-worker.ts and references/tail-workers.md ingest live request data.
Boundary markers: The Logger implementation in templates/logging-setup.ts and references/logging.md includes a redact method that uses regex patterns to strip sensitive keys like 'password', 'secret', and 'token' from logs.
Capability inventory: The skill uses fetch for network operations to send logs/alerts to external services (Slack, PagerDuty, Datadog) and executes the wrangler CLI via bash scripts.
Sanitization: CLI scripts include explicit validation of user-supplied arguments against strict charsets to mitigate injection risks.
[COMMAND_EXECUTION]: The skill provides developer utility scripts for log analysis and configuration management.
Evidence: scripts/analyze-logs.sh and scripts/setup-logging.sh execute system tools like wrangler, jq, grep, sed, and mktemp.
Safety Measures: The scripts use array-based command invocation and safe temporary file handling with backups to prevent command injection and accidental data loss.

cloudflare-workers-observability