ai-modcon/prism/prism-multi-agent-workflow
Claude Code and agent skill by ai-modcon/prism
- community
Checking that the SKILL.md file is still in the repository.
Work safely in the PRISM repo, which multiple coding agents edit in parallel. Use before committing, branching, pushing, deleting a branch/worktree, merging a stacked PR, or claiming "I'm on branch X." Triggers: "commit", "create a branch", "worktree", "delete branch", "clean up branches", "merge the PR", "stacked PR", "which branch am I on", "another agent edited this".
Install
npx vskill@latest install ai-modcon/prism/prism-multi-agent-workflowSKILL.md
PRISM — Multi-Agent Workflow
Default assumption: another agent is editing this clone right now. This repo is multi-agent by design (parallel agents on different PRs), so the bare working tree is a shared resource — another agent's git checkout / git stash / edit can silently change your state between turns. For the generic patterns, see the installed `multi-agent-git-hygiene` skill; this is the PRISM-specific drill.
Mandatory: work in an isolated worktree
- Call
EnterWorktreeat the start of any session that edits files or git state
→ creates .claude/worktrees/<name> on a fresh branch off origin/main.
- Spawning editing subagents? Pass
isolation: "worktree"to theAgenttool. - The bare clone (
/lus/flare/projects/ModCon/ngetty/BaseMM_PRISM) is
read-only — never commit or branch-switch there.
Mandatory: verify branch before you claim it
Before saying "I created branch X" / "I'm on branch Y" / "I'll commit to Z":
If you're on a branch you didn't create, STOP — git reflog | head -10; another agent almost certainly moved you. Do not commit. Back up your edits to /tmp/ first, then sort out where they belong.
A git status that shows your earlier edits as "no modifications" is a drift signal, not a no-op. Grep the file for a distinctive token from your edit before trusting it.
GitHub PRs are the source of truth for in-flight work
- Before a non-trivial task:
gh pr list --state opento see what's already in
flight. Don't pick up something another agent has open.
- One agent = one branch = one PR. Don't pile unrelated changes onto a branch
another agent owns.
Two destructive traps
- Check open PRs before pruning anything. Always
gh pr list --state open
first; never delete a branch/worktree backing an open PR even if the local name looks stale. (PR #87's worktree was nuked this way — 2026-05-25.)
- Stacked PRs need base-retarget before squash-merge. Squash-merging a
stacked PR whose base is a deleted branch silently orphans the content. Retarget the child PR's base to main first. (PRs #80/#81 lost — 2026-05-25.)
Launching jobs from a worktree
PBS scripts generated from a worktree need a ${PBS_O_WORKDIR:-...} fallback and a correct LAUNCHER_PRISM_DIR for the venv tarball — a hardcoded bare-clone path fails in ~1s when a plan YAML doesn't exist there. See prism-launching-jobs.
See also
- CLAUDE.md "Concurrency" section.
- prism-launching-jobs.
- Generic installed skills:
multi-agent-git-hygiene,shell-quoting-traps.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingdocs/skills/prism-multi-agent-workflow/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/ai-modcon/prism/prism-multi-agent-workflow)