The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the @launchdarkly/mcp-server package, as defined in marketplace.json. This is an official resource used to provide the toolset for feature flag interaction.
[COMMAND_EXECUTION]: The workflow involves calling Model Context Protocol (MCP) tools such as list-flags and get-flag. These tools are restricted to read-only lookups by the instructions in SKILL.md, and the skill explicitly forbids any write operations (create, toggle, update, or delete).
[PROMPT_INJECTION]: The skill contains robust guardrails in the 'Scope Boundary' section, explicitly instructing the agent to refuse requests for flag modification or staleness analysis. It mandates a specific routing message when users inquire about flag removal, ensuring the agent does not improvise safety conclusions.
[PROMPT_INJECTION]: Assessment of Indirect Prompt Injection risk surface:
Ingestion points: Data ingested from list-flags and get-flag tool outputs (SKILL.md).
Boundary markers: Not explicitly defined in the skill instructions.
Capability inventory: Restricted to read-only operations and routing to other local skill files; no write-access, file-system manipulation, or arbitrary code execution capabilities are present.
Sanitization: No specific sanitization logic is described in the instructions, though risk is mitigated by the limited read-only capability set.

launchdarkly-flag-command