implement-spec
Dispatch delivery of an accepted spec, or one ticket the existing spec already states, through Herdr workers to one PR.
Skill by varelaseb
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/varelaseb-implement-spec
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/implement-spec
What's inside
- Skill name
implement-spec- Description
- Dispatch delivery of an accepted spec, or one ticket the existing spec already states, through Herdr workers to one PR. The lane owner's route for every repo change.
- License
- Apache-2.0
- Instructions
- 245 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- openai.yaml · agents/
SKILL.md (first lines)
# Implement Spec Be extremely concise. Sacrifice grammar for the sake of concision. No em-dashes or en-dashes. Ever. The owner dispatches; it never edits repo files itself. Deliver the complete spec or ticket on one branch and leave one PR ready for human review. Workers load their `agents/*.md` role contracts and use octo-lite skills as guidance. Workers are the whole mechanism: no role resolver, loop workflow, journal, or exact-role gate sits on top of them. ## Inputs Require: - canonical spec - spec issue - target root `AGENTS.md` A ticket whose behavior the existing spec on main already states needs no shaping or new acceptance: that spec's commit on main is the spec acceptance receipt, and the ticket is the spec issue (a spec change goes back to shaping). The primary issue is the one-node graph: Start step 1 files nothing, and only filing is skipped. Every worker step still runs: one implementer in its own worktree, one integrator, one reviewer. The primary issue stays open until merge. In a repo whose specs are Spec Chat's, a ticket that changes a screen its spec's mockup set does not draw goes back to shaping, which draws it and returns the spec to review, before any implementer starts. A screen that must differ from its accepted mockup goes back the same way before the build differs (spec `#delivery-mockup`). A processed human **spec acceptance** (page action **Accept spec**, formerly called Finish review) permits implementation dispatch under the reviewed canonical spec. Follow `spec-chat/skill/review-spec/SKILL.md` under `## Remote hosting lifecycle` for review resource lifetime. It does not accept the implementation PR, approve a merge, promote to preproduction, or approve live traffic. The tickets are a task graph, not ordered steps. The ready frontier contains every incomplete ticket whose blockers are complete. Keep Linear titles, descriptions, relations, assignees, and states aligned with reality throughout. Do not copy ticket or spec prose into agent messages. Send essential absolute source paths in the message body; artifact metadata alone is insufficient. The lane owner keeps its lane record current as the lane-record hook in `agents/orchestrator.md` and `spec/domains/octo-lite.spec.html#lane-record` define. Record the requested handoff in the existing brief. Manual QA readiness means a usable feature deployment, required dependencies, working account roles and starting scenarios, with a secure credential pointer and stated limitations. Do not replace that outcome with exhaustive agent proof. Required spec/target evidence still applies; human QA, code readiness and acceptance are separate. ## Start 1. After spec acceptance, file the dependency-linked tickets under the spec issue as [consent](../octo-lite-github/SKILL.md#consent) allows, else keep the graph in the local brief: one outcome each, spec anchors, real blockers only. Before filing, read the target's changelog convention (file, format, which changes get an entry, by default every change its users can see) from its `AGENTS.md` or `CLAUDE.md`, or a rule file one of them points to; octo-lite names no changelog path or format. When the lane makes a change it logs, give the entry to one ticket (spec `#delivery-changelog`). 2. Read the spec, issue, and enough of the graph to identify the first frontier. 3. When exploration would remove duplicated implementation work, spawn one optional exploration agent through `herdr-comms`. Give it write access only to the lane's notes directory, `$(octo-lite-state)/lanes/<lane record name>-notes/` (spec `#delivery-notes`). Its deliverable is concise Markdown notes there, linked from the relevant tickets. 4. Continue the draft PR the orchestrator opened while shaping, else create the integration branch from the target base and open one draft PR. Push, open PRs, and file issues only as [consent](../octo-lite-github/SKILL.md#consent) allows; otherwise the lane stays on local branches and ends at a local integration head. Put `Closes <issue-key>` lines in the PR body for the spec issue and every implementation ticket, and one changelog line: `Changelog: <entry title>`, or `Changelog: none: <reason>` when the lane makes no change the convention logs or the target declares no convention. 5. Before dispatch, run the spec validator, backslash-artifact scan, and tag-balance scan. Repo tests run in delivery, never during shaping. Run one short cross-source conflict check: does the spec contradict another canonical spec, merged main, or an open PR? ## Herdr workers Use the verified launch path from `herdr-comms`; inspect help only when that invocation is unknown or changed. Spawn every exploration, implementation, merge, review-fix worker through Herdr. Prefer background tabs and fill available capacity from the current frontier. Spawn every worker into the owner's space, `--workspace "$HERDR_WORKSPACE_ID"`, named `<role>-<ticket>` lowercase (`implementer-ann-64`; a lane-wide worker such as the integrator carries the primary issue key), per spec `#transport-spawn-name`. Close a worker before reusing its name. Herdr owns session selection. Spawn Claude workers (`-- claude`, CLI default model; brief via `herdr-say`); Codex only when the operator explicitly selects it. Keep the worker-specific nuance: every writer still gets its own worktree, and only one integrator writes the integration branch. For each ready ticket: 1. Create a dedicated worktree and branch from the current integration branch. 2. Mark the Linear ticket In Progress only when its worker starts. 3. Spawn a Herdr implementer in that worktree, under the `implementer` role contract. Prompt it with pointers to the ticket, spec, target `AGENTS.md`, optional research notes, integration branch, and relevant prior commits. Name the accepted mockup of each screen the ticket changes: its file in the spec's mockup set at the spec acceptance receipt's commit (spec `#delivery-mockup`). Cite each repo-wide rule the ticket bears on by its marked criterion's spec and anchor (spec `#roles-repo-wide-rule`); an `AGENTS.md` line is only a pointer to it. The brief of the ticket that owns the changelog entry names the convention's file, so its implementer writes the entry in the same PR. 4. The contract declares the skills available to it and the bounds it works in, so neither has to be restated here or kept in step by hand. A Claude worker loads it with `--agent implementer`; a Codex worker gets it as developer instructions from `herdr-spawn --role implementer`. The worker owns only that ticket, validates it, commits it, and returns the
From varelaseb/octo-lite, Apache-2.0.
More from varelaseb/octo-lite
- 0.561score
octo-lite-github
varelaseb
Use GitHub PRs, branches, reviews, and explicitly requested GitHub issue operations in octo-lite workflows, while respecting Linear-first tracking when the…
SkillAFreshness: AClaude Code +9 - 0.561score
grill-with-docs
varelaseb
Run a docs-backed grilling session for octo-lite issue shaping; stress-test a draft or rough Linear/GitHub issue against target repo specs, ADRs, code, repo…
SkillAFreshness: AClaude Code +9 - 0.561score
push
varelaseb
Push the current octo-lite branch and create or update the linked GitHub PR using the target repo's conventions and validation evidence.
SkillAFreshness: AClaude Code +9 - 0.561score
repo-health
varelaseb
Run the repo health pass, fired by hand, that files a repo's real architecture and cleanliness issues as ranked health tickets; and file review notes and…
SkillAFreshness: AClaude Code +9 - 0.558score
herdr-comms
varelaseb
Spawn Herdr agents and send them messages, over native Herdr.
SkillAFreshness: AClaude Code +9 - 0.521score
nodejs
varelaseb
Node.js runtime policy for octo-lite implementer work when durable repo signals show a Node.js project and the issue touches runtime scripts, modules, env…
SkillBFreshness: BClaude Code +9 - 0.521score
frontend-design
varelaseb
Apply production-quality frontend design practice during octo-lite implementer work when the shaped issue touches UI, styling, interaction, or visual polish.
SkillBFreshness: BClaude Code +9 - 0.518score
typescript
varelaseb
TypeScript policy for octo-lite implementer work when durable repo signals show TypeScript is in scope.
SkillBFreshness: BClaude Code +9 - 0.518score
octo-lite-debug
varelaseb
Diagnose octo-lite workflow blockers in local Codex, git, gh, validation, branch, and PR state.
SkillBFreshness: BClaude Code +9 - 0.518score
python
varelaseb
Python stack policy for octo-lite implementer work when durable repo signals show Python is in scope.
SkillBFreshness: BClaude Code +9 - 0.499score
tdd
varelaseb
Test-driven development.
SkillBFreshness: BClaude Code +9 - 0.474score
commit
varelaseb
Create a well-formed git commit from current changes using session history, issue scope, and repo conventions; use when asked to commit or finalize staged…
SkillCFreshness: CClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 28
- Last commit (this folder)
- Oct 22, 2026 (7 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/varelaseb/octo-lite
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from varelaseb
- 0.561score
repo-health
varelaseb
Run the repo health pass, fired by hand, that files a repo's real architecture and cleanliness issues as ranked health tickets; and file review notes and…
SkillAFreshness: AClaude Code +9 - 0.561score
push
varelaseb
Push the current octo-lite branch and create or update the linked GitHub PR using the target repo's conventions and validation evidence.
SkillAFreshness: AClaude Code +9 - 0.561score
octo-lite-github
varelaseb
Use GitHub PRs, branches, reviews, and explicitly requested GitHub issue operations in octo-lite workflows, while respecting Linear-first tracking when the…
SkillAFreshness: AClaude Code +9 - 0.561score
grill-with-docs
varelaseb
Run a docs-backed grilling session for octo-lite issue shaping; stress-test a draft or rough Linear/GitHub issue against target repo specs, ADRs, code, repo…
SkillAFreshness: AClaude Code +9 - 0.558score
herdr-comms
varelaseb
Spawn Herdr agents and send them messages, over native Herdr.
SkillAFreshness: AClaude Code +9 - 0.521score
nodejs
varelaseb
Node.js runtime policy for octo-lite implementer work when durable repo signals show a Node.js project and the issue touches runtime scripts, modules, env…
SkillBFreshness: BClaude Code +9
Cite this page
implement-spec skill. Agent Mart. https://agentmart.dev/skills/varelaseb-implement-spec (last updated 2026-10-24)