The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to read metadata from the ~/.secrets file. While it includes strong warnings against dumping the file or leaking its contents, accessing this sensitive path is a risk.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its core functionality, including running local repository scripts like scripts/ci-profile/export.mjs, executing tests via pnpm exec vitest, and hashing identifiers using node -e.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external data sources including Honeycomb trace spans, CI events, and run logs. This external data represents an attack surface for indirect prompt injection if an attacker can manipulate trace attributes or log content.
Ingestion points: External data from the Honeycomb API, PR workflow logs, and CI profiling event data.
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded prompts within the trace data being analyzed.
Capability inventory: The skill has access to shell execution (node, pnpm) and file system operations in a developer-oriented environment.
Sanitization: There is no evidence of sanitization or structural validation for the external trace attributes before they are processed by the agent.
[DYNAMIC_EXECUTION]: The skill uses the node -e command to execute a JavaScript snippet at runtime to hash run IDs. Although the variables are passed securely as command arguments (process.argv[1]), the use of dynamic execution is noted.

honeycomb-tracing