Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/kavanaghpatrick/ralph-parallel/mergegit clone --depth 1 https://github.com/kavanaghpatrick/ralph-parallelWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00008 | $0.01607 |
| Opus 5 | $0.00004 | $0.00804 |
| Sonnet 5 | $0.00002 | $0.00321 |
| Haiku 4.5 | $0.00001 | $0.00161 |
Grade A, and why
merge scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 205 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Merge
Integrates results from parallel Agent Teams execution back into a cohesive state.
Note: For file-ownership strategy, this command is optional — it performs a consistency check only. Dispatch handles the full lifecycle including cleanup. Merge is required only for worktree strategy.
Parse Arguments
From $ARGUMENTS, extract:
- spec-name: Optional spec name (defaults to active spec)
- --abort: Cancel merge, restore pre-dispatch state
- --continue: Continue a merge that was paused for conflict resolution
Step 1: Load Dispatch State
1. Resolve spec (same logic as dispatch.md Step 1)
2. Read specs/$specName/.dispatch-state.json
- Error if missing: "No dispatch state found. Run /ralph-parallel:dispatch first."
3. Validate status field:
- "dispatched" → proceed with merge
- "merged" → "Already merged. Nothing to do."
- "superseded" → "This dispatch was superseded by a newer one. Nothing to merge."
- "merging" + no --continue → "Merge in progress. Use --continue to resume or --abort to cancel."
- "stale" → "This dispatch is stale (team lost at $staleSince). Run /ralph-parallel:dispatch to re-dispatch or /ralph-parallel:dispatch --abort to cancel."
Step 2: Verify Completion
Check that all dispatched groups have completed their work.
Completion Check:
1. FOR EACH group in dispatch state:
a. Check tasks.md — are all group's tasks marked [x]?
b. Check git log — are expected commits present?
c. If worktree strategy: check worktree branch exists and has commits
2. Report status:
- All complete → proceed to merge
- Partial → show which groups are incomplete:
"Group 'api-layer': 2/3 tasks complete (missing: 1.4)"
"Group 'data-layer': not started"
- Ask user: "Some groups incomplete. Merge anyway? (partial results)"
Step 3: File-Ownership Verification (Optional)
When strategy is file-ownership:
File-Ownership Verification (consistency check — dispatch already completed execution):
Since all teammates worked in the same directory with file ownership,
there's no git merge needed. Instead, verify consistency:
1. CHECK: All owned files were only modified by their assigned teammate
- For each group's ownedFiles:
- git log --oneline -- $file (check commit authors/messages)
- Verify commits come from expected group
- If file modified by wrong group: WARN "File $file was modified by
group '$other' but owned by '$owner'"
2. CHECK: No merge conflicts in working tree
- git status --porcelain
- If uncommitted changes: WARN and list them
3. CHECK: Build verification
a. Read qualityCommands from .dispatch-state.json
b. Run qualityCommands.build (if available)
c. Run qualityCommands.test (if available)
d. Run qualityCommands.lint (if available)
e. Collect pass/fail per command
f. If ANY fail: report specific failures with output, do NOT mark as merged
4. UPDATE: Mark dispatch as merged
- Update .dispatch-state.json: status = "merged"
- Update tasks.md: ensure all completed tasks marked [x]
- Update .progress.md with parallel execution summary
- Note: session-setup.sh will detect "merged" status and restore
gc.auto on the next session start
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 205 lines · 0 tokens per session scan A ec86a7a5e4d5
merge is a command published in the GitHub repository kavanaghpatrick/ralph-parallel (7 stars, last pushed 5mo ago), licensed MIT. It adds 8 tokens to every session and 1,607 once invoked, about $0.0000 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
git
Git operations with intelligent commit messages and workflow optimization.
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.