
Blitz
- 27 installs
- Updated January 1, 1970
- cygnusfear/agent-skills
Orchestrates parallel multi-issue sprints by spinning up git worktrees and delegating each issue to its own agent, then merging sequentially.
About
Blitz is a Claude Code skill that parallelizes multi-issue sprints using git worktrees and parallel agents. A solo builder reaches for it when they have three or more independent tasks that could be worked on at the same time, and it handles the whole flow from issue triage through delegating each to its own agent to a safe sequential merge.
- Runs 3+ independent issues concurrently
- Uses git worktrees for isolation
- Sequential merge to avoid conflicts
Blitz by the numbers
- 27 all-time installs (skills.sh)
- Ranked #9,535 of 16,556 AI & Agent Building skills by installs in the Skillselion catalog
- Data as of Jul 29, 2026 (Skillselion catalog sync)
npx skills add https://github.com/cygnusfear/agent-skills --skill blitzAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Installs | 27 |
|---|---|
| Last updated | January 1, 1970 |
| Repository | cygnusfear/agent-skills ↗ |
What it does
Orchestrates parallel multi-issue sprints by spinning up git worktrees and delegating each issue to its own agent, then merging sequentially.
Who is it for?
Tackling several independent issues concurrently
Skip if: A single linear task
Files
The Blitz: Parallel Worktree + Agent Workflow
Parallelizes multi-issue sprints by running independent agents in isolated git worktrees. Each agent implements their issue, self-reviews to 10/10 with 100% issue coverage, then changes are sequentially merged to avoid conflicts. Herding 🐲.
Platform-agnostic. Works with GitHub (ghCLI), Forgejo (API/teaCLI), or localtktickets.
See skills/obsidian-plan-wiki/references/platform-detection.md for detection rules and command mapping.⚠️ MANDATORY: 100% Issue Coverage Per Agent
Every agent MUST implement 100% of their assigned issue's requirements before their PR can be merged.
- Each agent receives COMPLETE issue requirements (extracted from issue/task)
- Review must verify ALL requirements are implemented
- Coverage < 100% = agent sent back to complete the work
- No PR merges until all requirements from the issue are addressed
Prerequisites
Required Tools:
- Platform CLI —
ghfor GitHub,tea/API for Forgejo, ortkfor local tickets - Git with worktree support (2.5+)
teamstool for parallel worker delegation
Required Skills:
4-step-program- Guides agents through fix-review-iterate-present loopcode-reviewer- Self-review to 10/10 quality gatedelphi- Parallel oracles for triage decisions (optional, for ambiguous triage)
Workflow Phases
Phase 1: Issue Triage
For ambiguous decisions on which issues to tackle, use the delphi skill:
Invoke Delphi: "Audit these open issues. For each, recommend: close (complete), fix (actionable), or defer (blocked)."Interpreting Delphi Results:
- Unanimous agreement → Act on recommendation
- 2/3 agreement → Lean toward majority, investigate minority view
- Full divergence → Need more context; investigate manually
Close complete issues immediately:
- GitHub:
gh issue close 1 2 3 --comment "Complete per Delphi audit" - Forgejo: Forgejo API
PATCH /repos/{owner}/{repo}/issues/{index}with"state": "closed" - Local tk:
tk close <id>for each completed task
For clear-cut issue lists, skip Delphi and proceed directly to Phase 2.
Phase 2: Worktree Setup
Create one worktree per fixable issue from main:
git worktree add .worktrees/<slug> -b fix/<slug> mainBranch Naming: fix/<descriptive-slug> or feat/<descriptive-slug>
Example setup for 4 issues:
git worktree add .worktrees/test-isolation -b fix/test-isolation main
git worktree add .worktrees/config-theater -b fix/config-theater main
git worktree add .worktrees/wire-salience -b fix/wire-salience main
git worktree add .worktrees/testing-quality -b fix/testing-quality mainWhy Worktrees:
- Complete filesystem isolation per agent
- No stash/checkout conflicts
- Agents work truly in parallel
- Each has independent node_modules, build artifacts
Phase 3: Delegate to Parallel Agents
Delegate to workers using teams. Each worker needs:
1. Working directory (absolute path to worktree) 2. Issue context (number, description, acceptance criteria) 3. COMPLETE list of ALL requirements from the issue (extracted via platform — see above) 4. Explicit instruction to use 4-step-program skill
CRITICAL: Before delegating, extract ALL requirements from each issue:
- GitHub:
gh issue view <number> - Forgejo:
tea issue view <number>or Forgejo API - Local tk:
tk show <id>
List EVERY requirement, acceptance criterion, and edge case in the agent prompt.
Agent Prompt Template (adapt to platform):
<details><summary><strong>GitHub variant</strong></summary>
Working directory: /absolute/path/to/.worktrees/<slug>
Issue: #<number> - <title>
**ALL REQUIREMENTS FROM ISSUE (100% must be implemented):**
1. [Requirement 1 from issue]
2. [Requirement 2 from issue]
3. [Requirement 3 from issue]
... (list ALL of them)
Use the 4-step-program skill to:
1. Implement ALL the above requirements (100% coverage required)
2. Run tests, verify passing
3. Create PR with `gh pr create` — **MUST include `Closes #<issue-number>` in body**
4. Self-review using code-reviewer skill (which will verify 100% coverage)
5. POST review to GitHub with `gh api`
**PR MUST include:**
- `Closes #<issue-number>` to auto-close the issue on merge
- "Related Issues" section in PR body
- Verify with `gh pr view --json closingIssuesReferences`
Do not return until you achieve 10/10 review score WITH 100% of issue requirements implemented AND issue properly linked.</details>
<details><summary><strong>Forgejo variant</strong></summary>
Working directory: /absolute/path/to/.worktrees/<slug>
Issue: #<number> - <title>
**ALL REQUIREMENTS FROM ISSUE (100% must be implemented):**
1. [Requirement 1 from issue]
2. [Requirement 2 from issue]
3. [Requirement 3 from issue]
... (list ALL of them)
Use the 4-step-program skill to:
1. Implement ALL the above requirements (100% coverage required)
2. Run tests, verify passing
3. Create PR with Forgejo API or `tea pr create` — **MUST include `Closes #<issue-number>` in body**
4. Self-review using code-reviewer skill (which will verify 100% coverage)
5. POST review to Forgejo via API
**PR MUST include:**
- `Closes #<issue-number>` to auto-close the issue on merge
- "Related Issues" section in PR body
Do not return until you achieve 10/10 review score WITH 100% of issue requirements implemented AND issue properly linked.</details>
<details><summary><strong>Local tk variant</strong></summary>
Working directory: /absolute/path/to/.worktrees/<slug>
Task: <tk-id> - <title>
**ALL REQUIREMENTS FROM TASK (100% must be implemented):**
1. [Requirement 1 from task]
2. [Requirement 2 from task]
3. [Requirement 3 from task]
... (list ALL of them)
Use the 4-step-program skill to:
1. Implement ALL the above requirements (100% coverage required)
2. Run tests, verify passing
3. Commit all changes — no PR needed, merge happens locally
4. Self-review using code-reviewer skill (which will verify 100% coverage)
5. Post review as a ticket: `todos_oneshot(title: "Review: <branch>", tags: "review")`
Do not return until you achieve 10/10 review score WITH 100% of task requirements implemented.</details>
CRITICAL: Agents must POST reviews to the platform, not just print them:
- GitHub:
gh api repos/OWNER/REPO/pulls/NUMBER/reviews -f body="..." -f event="COMMENT" - Forgejo:
POST /repos/{owner}/{repo}/pulls/{index}/reviewsvia Forgejo API - Local tk:
todos_oneshot(title: "Review: <branch>", tags: "review")— review lives as a ticket
Launch workers in parallel using teams:
teams(action: 'delegate', tasks: [
{text: '<agent prompt for issue 1>', assignee: 'fix-issue-1'},
{text: '<agent prompt for issue 2>', assignee: 'fix-issue-2'},
{text: '<agent prompt for issue 3>', assignee: 'fix-issue-3'}
])Phase 4: Review Iteration Loop
Monitor each PR/branch review status:
- GitHub:
gh pr view <NUMBER> --json reviews --jq '.reviews[-1].body' - Forgejo: Forgejo API
GET /repos/{owner}/{repo}/pulls/{index}/reviews - Local tk:
tk show <review-ticket-id>— check review ticket notes
TWO gates must pass for each PR:
GATE 1: 100% Issue Coverage
- Verify ALL requirements from original issue are implemented
- If ANY requirement is missing → Resume agent with missing requirements
GATE 2: 10/10 Review Quality
- Zero suggestions in review
- All verification commands pass
If coverage < 100%: Resume the agent with specific missing requirements:
PR #<NUMBER> coverage: 80% (4 of 5 requirements).
Missing requirement: [Requirement 5 from issue - specific text]
Implement this requirement and re-review.If score < 10/10 (but coverage 100%): Resume the agent with specific feedback:
PR #<NUMBER> has 100% coverage but scored 8/10. Issues:
- <specific issue 1>
- <specific issue 2>
Fix these issues and re-review.10/10 + 100% Coverage Criteria:
- ALL requirements from original issue implemented
- All functionality working
- Tests pass
- No obvious bugs or security issues
- Code follows project conventions
- Documentation updated if needed
Phase 4.5: FINAL COVERAGE GATE (Before Merge)
MANDATORY: Before merging ANY PR, perform LINE-BY-LINE requirement verification.
For EACH PR ready to merge:
Step 1: Extract ALL Requirements
GitHub:
gh issue view <issue-number> --json body --jq '.body' | grep -E "^\- \["
gh issue view <issue-number>Forgejo:
# Via Forgejo API: GET /repos/{owner}/{repo}/issues/{index}
tea issue view <issue-number>Local tk:
tk show <task-id>Step 2: Create Line-by-Line Table
MANDATORY for each PR:
## Issue #X - Full Requirements Check
| Requirement | PR Status | Evidence |
|-------------|-----------|----------|
| [exact text from issue] | ✅ | `file.cs:line` - [implementation] |
| [exact text from issue] | ❌ MISSING | Not found in PR |
| [exact text from issue] | ⚠️ PARTIAL | `file.cs:line` - [what's missing] |
| [exact text from issue] | ⚠️ MANUAL | Requires Unity Editor |Step 3: Honest Assessment
**Honest Assessment**:
- Coverage: X% (Y of Z requirements fully implemented)
- Missing: [list specific items]
- Partial: [list items and what's missing]
- Manual: [list items needing editor/runtime]FINAL GATE DECISION:
| Coverage | Action |
|---|---|
| 100% | ✅ Proceed to Phase 5 (Merge) |
| < 100% | ❌ DO NOT MERGE - Resume agent |
If Final Coverage < 100%:
Resume agent: "FINAL COVERAGE GATE FAILED for PR #<NUMBER>.
Issue #X - Full Requirements Check:
| Requirement | Status | Evidence |
|-------------|--------|----------|
| MeshDeformer component created | ✅ | MeshDeformer.cs |
| Create scene GameObject | ❌ MISSING | No scene modification |
| Cache hit rate >90% | ⚠️ MANUAL | Requires runtime profiler |
Honest Assessment:
- Coverage: 85% (11 of 13 requirements)
- Missing: scene GameObject creation
- Partial: none
- Manual: cache hit rate verification
Implement ALL items marked ❌. Items marked ⚠️ MANUAL that CAN be automated via mcp-unity MUST be automated.
Do not return until 100% coverage."→ Loop back to Phase 4 (Review Iteration)
Phase 5: Sequential Squash-Merge + Rebase
Merge PRs one at a time. Order by dependency (infrastructure first).
Before merging, verify issue linking:
- GitHub:
gh pr view <NUMBER> --json closingIssuesReferences --jq '.closingIssuesReferences[].number'— if empty, send agent back to fix - Forgejo: Check PR body for
Closes #Xkeywords - Local tk:
tk show <id>— verify deps/links are set
For each PR/branch:
GitHub:
# 1. Squash merge — issues auto-close on merge
gh pr merge <NUMBER> --squash --delete-branch
# 2. Update local main
git checkout main && git pull
# 3. Rebase next PR onto updated main
cd .worktrees/<next-slug>
git fetch origin main
git rebase origin/main
git push --force-with-leaseForgejo:
# 1. Merge via Forgejo API: POST /repos/{owner}/{repo}/pulls/{index}/merge
# with "Do": "squash"
# 2-3. Same rebase workflow as aboveLocal tk:
# 1. Fast-forward merge locally
git checkout main
git merge --ff-only fix/<slug>
# 2. Close the task
tk close <task-id>
# 3. Rebase next branch onto updated main
cd .worktrees/<next-slug>
git rebase mainWhy This Order:
- Squash merge (or ff-only for local) keeps main history linear
- Rebasing before merge prevents conflicts
- Sequential merging catches integration issues early
--force-with-leaseprevents overwriting others' work (remote platforms)
Handling Conflicts:
git rebase origin/main
# If conflicts:
# 1. Fix conflicts in affected files
# 2. git add <fixed-files>
# 3. git rebase --continue
# 4. git push --force-with-leasePhase 6: Cleanup
After all branches merge:
# Remove worktrees
git worktree remove .worktrees/<slug> # Repeat for each
# Delete local branches
git branch -D fix/<slug> # Repeat for each
# Sync main (remote platforms only)
git checkout main && git pull
# Verify clean state
git worktree list # Should show only main
git branch # Should show only mainQuick Reference
See references/commands.md for complete command reference. See references/pitfalls.md for common issues and solutions.
Mermaid Diagrams in Blitz PRs
Each agent's PR/commit SHOULD include Mermaid diagrams when the change warrants visualization.
When Agents Should Add Diagrams
| Change Type | Diagram |
|---|---|
| Flow change | flowchart before/after |
| API modification | sequenceDiagram |
| State handling | stateDiagram-v2 |
| Architecture change | flowchart with subgraphs |
Agent Delegation Should Include
When delegating to agents, add to the prompt:
If your changes involve flow modifications, state changes, or API interactions,
include a Mermaid diagram in the PR body (or commit message / review ticket) showing the new behavior.Example PR with Diagram
````markdown
Summary
Fixed race condition in WebSocket reconnection.
Before/After
flowchart LR
subgraph Before
A1[Disconnect] --> B1[Reconnect]
B1 --> C1[Duplicate handlers]
end
subgraph After
A2[Disconnect] --> B2[Cleanup handlers]
B2 --> C2[Reconnect]
C2 --> D2[Single handler]
endRelated Issues
- Closes #45 - WebSocket reconnection bug
````
---
Checklist Summary
1. [ ] Triage issues (use delphi if ambiguous) 2. [ ] Extract ALL requirements from each issue (platform: gh issue view / tea issue view / tk show) 3. [ ] Create worktrees for each fixable issue 4. [ ] Launch parallel agents with 4-step-program including complete requirement lists 5. [ ] Monitor and iterate until all PRs/branches hit 100% issue coverage AND 10/10 6. [ ] FINAL COVERAGE GATE: Re-verify 100% coverage before each merge 7. [ ] Sequential merge with rebase between (only after gate passes) 8. [ ] Cleanup worktrees and branches 9. [ ] PRs/commits include Mermaid diagrams where helpful
Blitz Command Reference
Platform-agnostic. Commands vary by platform. See skills/obsidian-plan-wiki/references/platform-detection.md for detection rules.Worktree Commands
# Create worktree with new branch from main
git worktree add .worktrees/NAME -b fix/NAME main
# List all worktrees
git worktree list
# Remove a worktree
git worktree remove .worktrees/NAME
# Clean stale worktree references (if manually deleted)
git worktree pruneIssue / Task Management
GitHub
gh issue close 1 2 3 --comment "Complete per audit"
gh issue view NUMBER
gh issue list --state openForgejo
tea issue view NUMBER
tea issue list
# Or via Forgejo API:
# PATCH /repos/{owner}/{repo}/issues/{index} — close
# GET /repos/{owner}/{repo}/issues/{index} — viewLocal tk
tk close <id>
tk show <id>
tk lsPR / Merge Commands
GitHub
# Create PR from current branch
gh pr create --title "TITLE" --body "BODY"
# View PR with reviews
gh pr view NUMBER --json reviews
# Get last review body
gh pr view NUMBER --json reviews --jq '.reviews[-1].body'
# Post review
gh api repos/OWNER/REPO/pulls/NUMBER/reviews \
-f body="REVIEW_CONTENT" -f event="COMMENT"
# Squash merge and delete branch
gh pr merge NUMBER --squash --delete-branchForgejo
# Create PR
tea pr create --title "TITLE" --description "BODY"
# Or via API: POST /repos/{owner}/{repo}/pulls
# View PR reviews
# GET /repos/{owner}/{repo}/pulls/{index}/reviews
# Post review
# POST /repos/{owner}/{repo}/pulls/{index}/reviews
# Merge PR
# POST /repos/{owner}/{repo}/pulls/{index}/merge with "Do": "squash"Local tk
# No PRs — merge locally
git checkout main
git merge --ff-only fix/NAME
# Post review as ticket
# todos_oneshot(title: "Review: <branch>", tags: "review")
# Close task after merge
tk close <id>Rebase Workflow
# Fetch latest main
git fetch origin main
# Rebase current branch onto main
git rebase origin/main
# Force push after rebase (safe version — remote platforms only)
git push --force-with-lease
# If conflicts during rebase:
# 1. Fix conflicts
git add <fixed-files>
git rebase --continue
# Or abort:
git rebase --abortBranch Cleanup
# Delete local branch
git branch -D fix/NAME
# Sync main after merges (remote platforms)
git checkout main && git pull
# Verify clean state
git worktree list
git branchProcess Monitoring
# Check for orphaned processes
ps aux | grep -E "(node|cargo|npm)"
# Kill specific process
kill <PID>Blitz Pitfalls and Gotchas
Agent Resume Errors
Agents cannot resume after API disconnection. If an agent fails mid-task:
1. Check its worktree state:
cd .worktrees/NAME && git status2. Spawn fresh agent with context of what's already done 3. The new agent picks up from the current state
Reviews as Comments vs Formal Reviews
How agent-posted reviews work depends on the platform:
GitHub:
gh apiposts reviews as PR comments, not GitHub's formal review system- Won't trigger "Changes requested" status or block merge via branch protection
- Fine for self-review workflow
- If formal reviews needed: use the GitHub web UI or different API endpoint
Forgejo:
- API-posted reviews behave similarly to GitHub
- Use
POST /repos/{owner}/{repo}/pulls/{index}/reviewsfor formal reviews - Comment-style reviews won't block merge
Local tk:
- Reviews are posted as tickets via
todos_oneshot(title: "Review: <branch>", tags: "review") - No merge protection — review is advisory
- Review ticket should be closed when agent achieves 10/10
Worktree Directory Confusion
Agents may drift to wrong directory during execution.
Prevention:
- Always specify absolute paths in prompts
- Verify with
pwdin agent's first command - Use
cd /absolute/path &&prefix for safety
Example prompt prefix:
Working directory: /Users/name/project/.worktrees/feature-x
First, verify: cd /Users/name/project/.worktrees/feature-x && pwdBackground Processes Accumulating
Long-running agents spawn background processes (dev servers, watchers, etc.).
Monitor with:
ps aux | grep -E "(node|cargo|npm)"Clean up orphans after workflow completes.
Stale Worktree References
If worktree directory was deleted manually (rm -rf instead of git worktree remove):
git worktree prune # Cleans stale referencesMerge Conflicts During Rebase
When rebasing onto updated main causes conflicts:
# During rebase
git rebase origin/main
# If conflicts appear:
# 1. Open conflicted files, resolve conflicts
# 2. Stage resolved files
git add <resolved-files>
# 3. Continue rebase
git rebase --continue
# 4. Force push (safe)
git push --force-with-leaseIf too messy: Abort and restart
git rebase --abort
git reset --hard origin/fix/NAME # Reset to remote stateAgent Quality Loop Stuck
If an agent keeps scoring below 10/10 after multiple iterations:
1. Review the feedback being given - is it actionable? 2. Check if the issue is actually fixable (may need deferral) 3. Consider human intervention for complex edge cases 4. Document the blocker and move on
Worktree Already Exists
fatal: '.worktrees/NAME' already existsSolutions:
- Use different slug name
- Remove existing:
git worktree remove .worktrees/NAME - Or prune if stale:
git worktree prune
Branch Already Exists
fatal: a branch named 'fix/NAME' already existsSolutions:
- Delete the branch:
git branch -D fix/NAME - Or use different branch name
- Check if work already exists on that branch