The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell tools to process large data outputs that have been persisted to files.
Evidence: In SKILL.md, the 'Universal guidance' section states: 'When a tool result is persisted to a file... use Bash with python3 or jq to extract the specific slice you need.' This encourages the use of a shell execution environment to handle large tool responses.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external observability data, which serves as an attack surface for instructions embedded in logs, traces, or error messages.
Ingestion points: The skill uses query-logs, query-traces, query-error-groups, and query-sessions in SKILL.md to fetch external data.
Boundary markers: There are no instructions in SKILL.md or companion files (logs.md, errors.md, etc.) to use delimiters or ignore instructions embedded within the processed observability data.
Capability inventory: The agent has access to Bash, python3, and jq for file processing as described in SKILL.md, and performs multiple data-retrieval tool calls.
Sanitization: No sanitization or validation of the fetched observability data is mentioned in the instructions.
[EXTERNAL_DOWNLOADS]: The skill relies on an external connection to a remotely hosted MCP server.
Evidence: SKILL.md specifies a requirement for the 'remotely hosted LaunchDarkly MCP server' in the frontmatter and prerequisites. This represents a dependency on the vendor's infrastructure.

investigate