audit-orphan-runs
Identifies abandoned worktrees and stale runs (no events for >2h, status=running) to flag for cleanup.
Conducts focused research on a topic by reading project files, searching memory, and synthesizing findings into a structured report.
$ npx -y skills add Samuel0101010/wisp-orchestrator --skill deep-research --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/deep-researchContext preview
The summary Claude sees to decide when to auto-load this skill.
Conducts focused research on a topic by reading project files, searching memory, and synthesizing findings into a structured report.
name: deep-research description: Conducts focused research on a topic by reading project files, searching memory, and synthesizing findings into a structured report. model: sonnet allowed-tools: ["Read", "Grep", "Glob", "WebFetch"] argument-hint: "<topic>" timeout-ms: 300000
You are a research specialist. Given the topic in the user message: 1. Explore the codebase via Read/Grep/Glob to gather concrete file references 2. If WebFetch is available, fetch up to 3 relevant external sources 3. Synthesize findings into:
Output only the report. No preamble.
Visual team-builder, plan-as-artifact, and live execution graph for autonomous Claude Code agent crews. Spawn a 3-role team, generate a DAG plan, run for hours, watch it ship in your browser.
Repo: Samuel0101010/wisp-orchestrator
Identifies abandoned worktrees and stale runs (no events for >2h, status=running) to flag for cleanup.
Drafts a docs/solutions/YYYY-MM-DD-<slug>.md entry for a recently completed run, capturing problem/solution/lessons.
Working discipline for code-writing agents — read before edit, smallest correct change, run the gates before claiming done. Injected into builder roles' system…
Self-diagnostics for the harness — checks DB integrity, MCP wiring, claude binary availability, and recent run failures.
Visual + interaction quality bar for frontend roles — states, responsiveness, accessibility, consistency. Injected into frontend roles' system prompts.
Verification method for QA roles — execute, don't assume; report evidence, not impressions. Injected into QA roles' system prompts.