Finishing a Development Branch
obra/superpowers
Walks the last step of a branch: confirm tests pass, detect the git environment, ask how to integrate, carry out your choice and clean up the worktree.
Coordinator workflow for multi-phase implementation across workmux worktrees.
$ npx skills add raine/consult-llm --skill phased-implement -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install raine/consult-llm phased-implement --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/raine/consult-llm.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/phased-implement .claude/skills/phased-implement && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "phased-implement" agent skill from https://github.com/raine/consult-llm/tree/main/skills/phased-implement into .claude/skills/phased-implement/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "phased-implement", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/raine/consult-llm/tree/main/skills/phased-implementType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add raine/consult-llm --skill phased-implement -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install raine/consult-llm phased-implement --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/raine/consult-llm.git skills-src && mkdir -p .agents/skills && cp -r skills-src/skills/phased-implement .agents/skills/phased-implement && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "phased-implement" agent skill from https://github.com/raine/consult-llm/tree/main/skills/phased-implement into .agents/skills/phased-implement/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "phased-implement", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add raine/consult-llm --skill phased-implement -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install raine/consult-llm phased-implement --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/raine/consult-llm.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/skills/phased-implement .cursor/skills/phased-implement && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "phased-implement" agent skill from https://github.com/raine/consult-llm/tree/main/skills/phased-implement into .cursor/skills/phased-implement/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "phased-implement", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/raine/consult-llm.git --path skills/phased-implement--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add raine/consult-llm --skill phased-implement -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install raine/consult-llm phased-implement --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/raine/consult-llm.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/skills/phased-implement .gemini/skills/phased-implement && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "phased-implement" agent skill from https://github.com/raine/consult-llm/tree/main/skills/phased-implement into .gemini/skills/phased-implement/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "phased-implement", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install raine/consult-llm phased-implementInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add raine/consult-llm --skill phased-implement -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/raine/consult-llm.git skills-src && mkdir -p .github/skills && cp -r skills-src/skills/phased-implement .github/skills/phased-implement && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "phased-implement" agent skill from https://github.com/raine/consult-llm/tree/main/skills/phased-implement into .github/skills/phased-implement/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "phased-implement", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add raine/consult-llm --skill phased-implement -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install raine/consult-llm phased-implement --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/raine/consult-llm.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/skills/phased-implement .opencode/skills/phased-implement && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "phased-implement" agent skill from https://github.com/raine/consult-llm/tree/main/skills/phased-implement into .opencode/skills/phased-implement/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "phased-implement", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
phased-implementCoordinator workflow for multi-phase implementation across workmux worktrees.
Phased Implement is an agent skill from raine/consult-llm. Coordinator workflow for multi-phase implementation across workmux worktrees. Generates or loads a master plan, dispatches phase agents using presets, verifies sentinels, merges serially, and performs integration verification.
Its SKILL.md is about 4.2k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.
It sits in Development, covering Git worktrees. The repository describes itself as: Get a second opinion from another AI model. The licence is MIT.
6 steps, taken from the first numbered list in SKILL.md.
Read from SKILL.md and the folder at commit 69e3ecb. It shows what the files ask for, not the result of running them.
Pre-approves these tools, so the agent can use them without asking each time:
BashReadWriteGlobGrepFrom allowed-tools in the SKILL.md frontmatter.
Shell commands in SKILL.md call:
gitFrom the folder's file list and the shell code blocks in SKILL.md.
No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.
From URLs in SKILL.md, links to its own repository left out.
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Phased Implement loads about 4.2k tokens when it runs. Until then it costs about 61 tokens; SKILL.md has 1,615 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check noted patterns worth knowing about, such as sudo or a known installer.
allowed-tools: Bash, Read, Write, Glob, GrepAutomated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
The full file from raine/consult-llm at commit 69e3ecb, republished under its MIT licence (© raine). 1,615 words, ~4,164 tokens.
.claude/skills/phased-implement/SKILL.md (or your agent's skills folder).Coordinate multi-phase implementation across workmux worktrees. The coordinator owns the master plan, phase dispatch, merge order, ancestry checks, final integration validation, and final summary. The coordinator does not edit source files for feature work.
Do not use Claude Code built-in worktree features. Use workmux for worktree orchestration.
/implement.workmux done is not success. A phase succeeds only when its result sentinel reports status: success and the coordinator verifies it./merge --keep runs at a time.workmux wait --any can return after one transition while sibling phases also finished.workmux status on timeouts.history/<date>-phased-<slug>/ run directory so worktrees can read prompts and write sentinels.Arguments are $ARGUMENTS.
Parse these flags before starting:
--plan <path>: load an existing master plan instead of generating one.--integration-branch <branch>: branch that receives phase merges. Default: current branch.--preset light|standard|design|strict: default preset for generated phases.--validation <command>: final integration validation command.--reviewer <selector>: consult-llm selector for optional integration review. Repeatable.--reviewers <selector,selector>: comma-separated reviewer selectors.--integration-review none|auto|full: final integration review policy. Default: auto.Everything else is the requested multi-phase implementation.
Preset semantics:
light: established local pattern, focused validation, no external reviewstandard: bounded implementation with acceptance evidence and applicable
runtime exercisedesign: meaningful ownership, API, or architecture choice, with one
source-grounded technical-shape reviewstrict: real trust, persistence, protocol, migration, destructive,
concurrency, or FFI boundary, with boundary evidence and one final diff reviewPresets increase evidence, not planning or architecture.
Write coordinator artifacts under history/ using a dated run directory:
history/<YYYY-MM-DD>-phased-<slug>/
plan.md
prompts/
captures/
summary.mdRequired:
history/<run>/plan.mdhistory/<run>/captures/<phase-id>.result.mdhistory/<run>/summary.mdOptional:
Do not require ADRs or feedback ledgers.
Record the integration branch and start commit:
START_HEAD=$(git rev-parse HEAD)
INTEGRATION_BRANCH=<integration-branch or current branch>
git branch --show-current
git rev-parse --verify "$INTEGRATION_BRANCH"
git status --shortStop if unrelated uncommitted changes make orchestration unsafe. The integration branch must exist and must not be a detached HEAD.
Pick a topic slug and create the shared run directory:
PLAN_DIR="history/<YYYY-MM-DD>-phased-<slug>"
mkdir -p "$PLAN_DIR/prompts" "$PLAN_DIR/captures"history/ is gitignored and shared across workmux worktrees in this repo style. Do not commit files from this directory.
Track phase state in coordinator memory with these statuses: pending, working, done-unverified, merging, merged, failed, blocked.
If --plan <path> is provided, copy or write it to $PLAN_DIR/plan.md, read it, and validate it.
If no plan is provided, generate a master phased plan at $PLAN_DIR/plan.md. Gather source context first with Glob, Grep, and Read. The master plan should be scheduling-focused, not code-heavy. Assign phase presets by risk:
light for routine mechanical phases from a clear master planstandard for non-trivial phases with known approachdesign for uncertain, cross-module, or approach-heavy phasesstrict for public contracts, data formats, security-sensitive work, migrations, or high regression riskGenerated plans should mark non-trivial phases as at least standard.
The master plan must contain one YAML block and one brief per phase.
# <Feature> Master Plan
**Goal:** <one sentence>
**Integration branch:** <branch>
**Start commit:** <sha>
**Final validation:** `<command>`
```yaml
phases:
- id: api-contract
description: Define the public API contract for X.
depends_on: []
paths:
- "src/api/**"
- "tests/api/**"
preset: design
acceptance:
- "Given a valid request, when the API is called, then it returns the new response shape."
validation: npm test -- tests/api
```
## Phase briefs
### api-contract
**Intent:** why this phase exists
**Current problem:** what is wrong now
**Desired shape:** target behavior and boundaries
**Preserve:** behavior that must not change
**Avoid:** overreach and later-phase ownership
**Dependencies:** previous phase outputsRequired YAML fields per phase:
id: stable slug, unique across phasesdescription: one sentencedepends_on: list of phase idspaths: expected owned paths or globspreset: light, standard, design, or strictacceptance: list of acceptance criteriaOptional field:
validation: phase-specific validation commandBefore dispatch, validate the plan:
If validation fails, update the master plan directly and rerun validation.
Dispatch phases whose dependencies have succeeded and merged. Parallel phases may run at the same time when dependencies are satisfied and path ownership is safe.
Use the exact local workmux command syntax when running commands in this repo. Do not switch to Claude Code worktrees.
Loop:
Drain completed handles before spawning new work. Run Phase C and Phase D for every tracked handle that is already done before recomputing the ready set.
Compute the ready set: phases with status pending whose every dependency has status merged.
If the ready set is empty and no phase is working, done-unverified, or merging, proceed to final integration validation.
For each ready phase:
$PLAN_DIR/prompts/<phase-id>.md.working.workmux add <phase-id> -b --base "$INTEGRATION_BRANCH" -P "$PLAN_DIR/prompts/<phase-id>.md"Confirm newly spawned handles started with a bounded wait:
workmux wait <handles> --status working --timeout 120On timeout, inspect workmux status. A fast phase may have moved straight to done; treat done as valid. Mark a handle failed only if it is missing, exited unexpectedly, or is stuck before it starts.
Recompute the live working handle set before every wait. Exclude phases that are already done-unverified, merging, merged, failed, or blocked.
Wait for the next transition in bounded chunks:
workmux wait <working-handles> --any --timeout 300
workmux status <working-handles>If timeout occurs, inspect workmux status. If any handle is waiting, capture it, mark it blocked, and halt its dependents. Otherwise continue waiting. If a handle exited unexpectedly, mark it failed.
Treat done as unverified completion, not success. Set status to done-unverified, capture output, and run Phase C.
After every done handle has been verified and merged or marked failed, recompute the ready set. Skipping this drain can spawn dependents against a stale integration branch.
Halt logic:
failed or blocked, stop spawning new dependent phases.blocked with the blocker phase id.Write one prompt file per phase:
# Phase Agent Prompt: <phase-id>
You are implementing one phase in a workmux worktree.
## Hard rules
- Work only on this phase.
- Invoke `/implement` for this phase with the resolved preset and phase context.
- Preserve the phase boundaries, acceptance criteria, and dependencies below.
- Commit successful changes in the phase worktree.
- Write the result sentinel exactly as requested.
## Invoke
Run this implementation workflow:
`/implement --preset <preset> --parent-plan <master-plan-path> --validation '<phase-validation-command>' <phase description and acceptance>`
## Phase context
- Phase id: `<phase-id>`
- Description: <description>
- Paths: <paths>
- Preset: `<preset>`
- Master plan: `<path>`
- Dependencies: <dependencies>
## Acceptance criteria
- <criterion>
## Phase brief
<brief from master plan>
## Result sentinel
Write `<plan-dir>/captures/<phase-id>.result.md`:
```markdown
# Phase Result: <phase-id>
status: success | blocked | failed
phase_id: <phase-id>
preset: light | standard | design | strict
worktree: <workmux worktree name>
base_commit: <sha>
head_commit: <sha>
commit: <sha>
validation: <command>
validation_status: passed | failed | skipped
## Summary
- <what changed>
## Acceptance
- <criterion>: met | not met | unknown
## Files changed
- <path>
## Blockers
- <blocker or none>
The phase agent should write the sentinel as its final action and treat a missing sentinel as failure.
## Phase C: result verification
When a handle transitions to `done`, capture output before deciding whether it succeeded:
```bash
workmux capture <phase-id> > "$PLAN_DIR/captures/<phase-id>.tail"Then read $PLAN_DIR/captures/<phase-id>.result.md. For each completed phase, the coordinator verifies:
status is success.head_commit and commit are present.none.If the sentinel is missing, reports blocked or failed, or fails verification, do not merge that phase. Capture the workmux output, mark the phase failed or blocked, preserve the worktree, and halt dependents.
Merge successful phases serially into the integration branch. Before each merge, ensure the integration branch is checked out and current.
Use the existing merge skill in the phase worktree:
/merge --keepAt a high level:
Capture the phase tip from the sentinel before merge for diagnostics: head_commit.
Send /merge --keep to the phase worktree agent with workmux.
Wait for merge completion in bounded chunks so conflicts or prompts are detected:
workmux send <phase-id> "/merge --keep"
workmux wait <phase-id> --timeout 60
workmux status <phase-id>Repeat the wait/status cycle until the merge finishes, a bounded timeout is reached, or the handle enters waiting. If it enters waiting, capture output to $PLAN_DIR/captures/<phase-id>.merge.tail, keep the worktree, mark the phase blocked, and stop merging dependents.
Because /merge may rebase the phase before merging, read the worktree tip after /merge --keep returns. The pre-merge head_commit from the sentinel is not a stable ancestry token.
Verify the post-merge worktree tip is an ancestor of the integration branch:
POST_MERGE_TIP=$(workmux run <phase-id> -- git rev-parse HEAD)
git merge-base --is-ancestor "$POST_MERGE_TIP" <integration-branch>If ancestry verification passes, record POST_MERGE_TIP as the merged phase commit and remove the workmux worktree:
workmux remove <phase-id>If ancestry verification fails, capture output, keep the worktree, stop merging, and report the phase as blocked.
Do not use workmux remove before the merge and post-merge ancestry verification succeed.
The sentinel's pre-merge head_commit is for diagnostics only, not ancestry verification.
After each successful merge:
If a phase fails, keep unrelated ready phases running only when their dependencies and paths are unaffected. Otherwise stop and summarize.
When all phases merge:
--validation.Final integration review policy:
none: skip external integration review.auto: run full external integration review when any phase used design or strict, or when multiple phases changed shared contracts. Otherwise skip.full: always run full external integration review.For final integration review, load the consult-llm skill before calling consult-llm. Attach the master plan, phase sentinels, and relevant diffs or files. Use --task review, supplied reviewer selectors if present, quoted heredoc terminator __CONSULT_LLM_END__, and Bash timeout 600000.
Prompt shape:
Review the integrated result of this multi-phase implementation.
Check whether the merged phases satisfy the master plan acceptance criteria. Focus on cross-phase compatibility, public contracts, regression risk, edge cases, validation adequacy, and security where relevant.
Return only actionable must-fix or should-fix issues. If the result is acceptable, say so directly.Only auto-fix localized must-fix issues with one clear answer. If review finds design or scope issues, stop and summarize them.
Write a final summary under history/ and print it:
# Phased Implementation Summary: <feature>
status: success | blocked | failed
integration_branch: <branch>
start_commit: <sha>
end_commit: <sha>
master_plan: <path>
final_validation: <command>
final_validation_status: passed | failed | skipped
integration_review: none | skipped | passed | failed
## Phases
| Phase | Preset | Status | Commit | Sentinel |
| --- | --- | --- | --- | --- |
| <id> | <preset> | merged | <sha> | <path> |
## Acceptance
- <criterion>: met | not met | unknown
## Merges
- <phase>: merged and ancestry verified against post-merge tip `<post-merge-tip>`
## Blockers
- <blocker or none>
## Artifacts
- Prompts: `<plan-dir>/prompts/`
- Captures: `<plan-dir>/captures/`
## Next steps
- <only if needed>If phases failed or blocked, list preserved worktrees and the capture files to inspect. Report the final validation result, integration review result, merged phase commits, and blockers. If all phases merged and validation passed, state success plainly.
© raine, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Just SKILL.md in skills/phased-implement of raine/consult-llm.
Open the folder on GitHubat commit 69e3ecb
Phased Implement next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Phased Implement this skillraine/consult-llm | 139 | — | ~4.2k | Automated safety check: Notes | MIT | |
| Finishing a Development Branchobra/superpowers | 296k | 5 repos | ~1.9k | Automated safety check: Pass | MIT | |
| Migrate Core Code to Submodulestinyhumansai/openhuman | 42k | — | ~2.6k | Automated safety check: Pass | GPL-3.0 | |
| Finishing A Development Branchfarm-fe/farm | 5.6k | 34 repos | ~1.8k | Automated safety check: Pass | MIT | |
| Git Worktree Cleanuplobehub/lobehub | 83k | — | ~2.8k | Automated safety check: Pass | Custom licence | |
| Keep Codex Fastvibeforge1111/keep-codex-fast | 1.6k | — | ~3.1k | Automated safety check: Pass | MIT |
obra/superpowers
Walks the last step of a branch: confirm tests pass, detect the git environment, ask how to integrate, carry out your choice and clean up the worktree.
tinyhumansai/openhuman
Plans and carries out moving non-host-specific code and its tests from the OpenHuman core into vendored tiny submodule libraries, then releases the submodule and re-pins the host.
farm-fe/farm
A skill your agent uses when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for…
lobehub/lobehub
Audits stale Git worktrees and branches with a bundled script, classifies each one, and deletes only after you approve the exact candidates.
vibeforge1111/keep-codex-fast
A skill your agent uses when Codex feels slow or bloated, when local sessions/logs/worktrees/config have grown over time, or when a user wants safe maintenance for Codex Desktop/CLI state.
jamiepine/voicebox
Sorts a backlog of open pull requests into must-merge, candidate, superseded and deferred, writes a triage doc and works the merge loop before a release.
raine/consult-llm
Explicit workflow for one bounded implementation using source-grounded discovery, a walking slice, evidence-gated review, validation, and commit.
raine/consult-llm
Multiple LLMs collaboratively brainstorm solutions, building on each other's ideas across rounds.
raine/consult-llm
The agent brainstorms with a partner LLM in alternating turns, building on each other's ideas.
raine/consult-llm
Consult an external LLM with the user's query. An agent skill from raine/consult-llm.
raine/consult-llm
How to invoke the consult-llm CLI. An agent skill from raine/consult-llm.
raine/consult-llm
LLMs propose and critique approaches, agent moderates the debate and synthesizes the best solution, then implements.
Categories
Coordinator workflow for multi-phase implementation across workmux worktrees. Phased Implement is an agent skill from raine/consult-llm. Coordinator workflow for multi-phase implementation across workmux worktrees.
Phased Implement fits situations like: tasks that involve Git worktrees.
Run `npx skills add raine/consult-llm --skill phased-implement -a claude-code`. Or copy the skill folder (skills/phased-implement in raine/consult-llm) into .claude/skills/phased-implement in your project. Claude Code loads it when a task matches its description.
Run `npx skills add raine/consult-llm --skill phased-implement -a codex`. Or copy the skill folder (skills/phased-implement in raine/consult-llm) into .agents/skills/phased-implement in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add raine/consult-llm --skill phased-implement -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/phased-implement, .gemini/skills/phased-implement, .github/skills/phased-implement and .opencode/skills/phased-implement in your project.
Going by SKILL.md and its folder, Phased Implement needs the command-line tools its instructions call (git). Its frontmatter pre-approves these tools: Bash, Read, Write, Glob, Grep.
SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found notes only (pre-approves every shell command (allowed-tools: bash)), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.
Phased Implement is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 4.2k tokens (SKILL.md is roughly 17k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
Skills that share tags, products or a category with Phased Implement: Finishing a Development Branch (obra/superpowers, 296k stars), Migrate Core Code to Submodules (tinyhumansai/openhuman, 42k stars), Finishing A Development Branch (farm-fe/farm, 5.6k stars) and Git Worktree Cleanup (lobehub/lobehub, 83k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
raine (a GitHub user) maintains it in raine/consult-llm, which has 139 GitHub stars. The repository holds 12 skills in this directory. The repository was last updated on October 7, 2026.
Source: raine/consult-llm on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.