Skip to content

CLAUDE.md

Latest commit

 

History

History
32 lines (22 loc) · 2.45 KB

CLAUDE.md

File metadata and controls

32 lines (22 loc) · 2.45 KB
Raw

CLAUDE.md

Project-level instructions for Claude Code when working in the fusion-skills plugin. For the tool-agnostic guide (repo structure, skills ecosystem, usage without the plugin system), see AGENTS.md.

Plugin Hook Behavior

This plugin includes two hooks that run automatically:

  • UserPromptSubmit + PreToolUse: fusion-skill-router.sh detects Fusion workflow intent in the prompt, writes a short-lived marker file, and injects advisory context steering toward the workflows orchestrator skill. On later tool calls it emits a non-blocking reminder until the Skill tool is invoked.
  • PreToolUse (Skill): fusion-foundry-bridge.sh provides advisory cross-plugin routing between crowdstrike-falcon-fusion (workflows) and crowdstrike-falcon-foundry (Foundry app lifecycle).

Both hooks are advisory only — they always exit 0 and never block a user action or a skill invocation.

Counter-Rationalizations

The skills enforce discipline to prevent common failures. When you catch yourself thinking one of these, stop:

Thought Reality
"I'll write the YAML without searching actions" STOP. Run action_search.py first — action IDs are opaque catalog identifiers, only discoverable via API
"I'll use a placeholder ID for now" NEVER. Resolve every action ID before writing YAML. No PLACEHOLDER_* values
"version_constraint is optional" WRONG. Every action requires it: ~<major> of its semantic_version (~0 when none, e.g. Charlotte AI at 0.0.100)
"Validation can wait until deploy" NO. Authoring validates; deployment validates again as a pre-flight
"I'll deploy without releasing" INCOMPLETE. Workflows must be released before they can execute
"This is basically a Foundry app" CHECK. If it needs UI/functions/collections, route to foundry-skills instead

Skills Integration

  • Orchestration: The workflows skill is the entry point. It routes to authoring, deployment, and execution, and coordinates the full lifecycle for end-to-end requests.
  • Direct invocation: Each sub-skill can be invoked directly for a focused task (e.g., just validating an existing YAML file).
  • Shared auth: All Python scripts import credentials from common/scripts/auth.py (get_client() for Fusion, get_ngsiem_client() for lookup files).
  • Superpowers: If installed, superpowers planning/TDD skills MAY supplement the workflow but should not replace the orchestrator's routing.