refine
Take a backlog item from idea to approved spec via Clarify → Research → Spec.
Skill by charleshall888
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/charleshall888-refine
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
Use cases: research
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 and code. Checked Oct 26, 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/refine
What's inside
- Skill name
refine- Description
- Take a backlog item from idea to approved spec via Clarify → Research → Spec. Stops at spec; build takes it from there. Also preps for overnight.
- License
- MIT
- Instructions
- 59 lines in SKILL.md · view on GitHub
Files (6)
- SKILL.md
- clarify-critic.md · references/
- clarify.md · references/
- research-phase.md · references/
- size-pin.txt · references/
- specify.md · references/
SKILL.md (first lines)
# Refine
Clarify → Research → Spec. Ends at `status: refined` with a linked spec, ready for `/cortex-core:build`. Between phases, announce and continue without asking. Stop to ask only at the `<!-- pause: -->` markers here and in the references.
<!-- pause: refine-empty-topic-prompt question -->
Topic: $ARGUMENTS. Empty → ask for one first.
## 1. Start
```bash
cortex-refine start <input>
```
Use the fields it returns as they are. Keep `backend`, `lifecycle_slug`, and `backlog_filename_slug` for the whole run — every write-back needs them.
- `ready` → continue. `needs-slug` (Context B: no matching item) → make a short kebab slug, announce it, re-run with `--lifecycle-slug`. Exit 2 → several items match; the candidates are on stderr, let the user pick. Exit 70 → report the error and stop.
- `epic_research` / `epic_spec` set → read them as background, show any `warning`, add a `## Epic Reference` link section to `research.md` and a preamble note to `spec.md`. Never copy epic content in — it covers every child ticket.
- `resume`: `complete` → announce and skip to §5. Re-run only if the user asks; that overwrites the spec and resets `status: in_progress`. `research` → a spec exists without research: warn that overnight needs both, skip Clarify, run Research. `spec` → resume at Spec. `clarify` → start there.
## 2. Clarify
Follow `${CLAUDE_SKILL_DIR}/references/clarify.md`; carry its §5 outputs forward.
Write complexity and criticality back right away (Context A only). Every backlog write here uses this **3-arm routing** on `backend`: `cortex-backlog` → `cortex-update-item {backlog-filename-slug} --complexity {value} --criticality {value}`; `none` → skip and say so in one line; external → try per `backlog.instructions`, and show the values if it fails. Exit 2 → the ambiguous-slug rule in `${CLAUDE_SKILL_DIR}/../build/references/backlog-writeback.md`.
**Stop at `simple`** — no lifecycle needed. Say so and hand back to direct implementation (dev rule 4). Continue only at `moderate` or `complex`.
## 3. Research
Follow `${CLAUDE_SKILL_DIR}/references/research-phase.md`. Then re-check the tier against Clarify's rubric. Only if it changed: `cortex-lifecycle-event complexity-override --feature <feature> --from <old> --to <new> --reason "{tag}: <one line>"`. The reason points to the argument in `research.md`; the tag is optional, one of `reversibility:` / `exposure:` / `consequence:` / `other:`.
## 4. Spec
Reconcile first. The lifecycle state still holds pre-Clarify defaults, and specify.md §3b's critical-review check reads the reconciled values:
```bash
# Context A — reads values from backlog frontmatter
cortex-refine reconcile-clarify --backend {resolved} --lifecycle-slug {lifecycle-slug} --backlog-slug {backlog-filename-slug} --criticality-reason "{tag}: {why}" --tier-reason "{tag}: {why}"
# Context B — passes Clarify's computed values
cortex-refine reconcile-clarify --backend {resolved} --lifecycle-slug {lifecycle-slug} --complexity {value} --criticality {value} --criticality-reason "{tag}: {why}" --tier-reason "{tag}: {why}"
```
The `--*-reason` flags are optional one-liners reusing Clarify's reasoning. Omit them rather than use placeholders.
Then follow `${CLAUDE_SKILL_DIR}/references/specify.md` in full; its orchestrator-review path is `${CLAUDE_SKILL_DIR}/../build/references/orchestrator-review.md`.
The spec-approve verb does the approval write-back itself (same 3-arm routing). Pass it `--backend {resolved}`, `--backlog-file {backlog-filename-slug}` (`""` in Context B), `--spec-path cortex/lifecycle/{lifecycle-slug}/spec.md`, `--no-emit-transition` (refine stops at spec), and areas: `--areas a b` to set, `--clear-areas` to empty, omit to keep. Areas name the main subsystem changed (standard names: `overnight-runner`, `backlog`, `skills`, `lifecycle`, `hooks`, `report`, `tests`, `docs`); 4+ with no main one → clear.
## 5. Finish
```bash
cortex-lifecycle-stage-artifacts --phase refine --feature {lifecycle-slug} --commit-subject "Refine {feature}: research and spec"
```
Use `Refine {feature}: cancelled at spec approval` when `spec.md` is absent. `config_disabled` → show `message`. `nothing_staged` → nothing to do. `staged` → the verb committed; show `commit.sha`, or on `failed` show `commit.message` and stop.
Announce the item, lifecycle directory, artifacts, fields written (`complexity`, `criticality`, `status: refined`, `spec`, `areas`), and that `/cortex-core:build {lifecycle-slug}` is next.From charleshall888/cortex-command, MIT.
More from charleshall888/cortex-command
- 0.606score
overnight
charleshall888
Plan and launch an unattended overnight session — pick refined features, approve the plan, hand off to the runner.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
android-cli
charleshall888
Orchestrates Android development tasks including project creation, deployment, SDK management, and environment diagnostics using the `android` command-line…
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
build
charleshall888
Take a refined ticket from spec through plan, implement, review, and complete.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
critical-review
charleshall888
Adversarial review — 1–2 fresh reviewer agents on distinct angles, consolidated here.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
dev
charleshall888
Development entry point — routes a request to the right workflow.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
discovery
charleshall888
Investigate a topic, then decompose it into backlog tickets grouped by epic.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
research
charleshall888
Parallel research orchestrator — dispatches 1–6 agents across independent angles, synthesizes into research.md.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
morning-review
charleshall888
Walk the morning report after an overnight run — verify features, answer deferred questions, advance lifecycles, merge the PR, close tickets.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
harness-review
charleshall888
Evaluate overnight runner components against a curated inventory using a 4-question rubric to classify each as load-bearing, experiment-candidate, or…
SkillAFreshness: A★ 3Claude Code +9 - 0.602score
r8-analyzer
Google LLC
Analyzes Android build files and R8 keep rules to identify redundancies, broad package-wide rules, and rules that subsume library consumer keep rules.
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
requirements
charleshall888
Use /cortex-core:requirements to gather requirements or define project scope.
SkillAFreshness: A★ 3Claude Code +9 - 0.601score
writing-great-skills
charleshall888
Reference for writing and editing skills well — the vocabulary and principles that make a skill predictable.
SkillAFreshness: A★ 3Claude Code +9
Signals
- Stars
- 3
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 22, 2026 (1 month ago)
- Last release
- Oct 5, 2026 (24 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/charleshall888/cortex-command
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from charleshall888
- 0.606score
android-cli
charleshall888
Orchestrates Android development tasks including project creation, deployment, SDK management, and environment diagnostics using the `android` command-line…
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
build
charleshall888
Take a refined ticket from spec through plan, implement, review, and complete.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
critical-review
charleshall888
Adversarial review — 1–2 fresh reviewer agents on distinct angles, consolidated here.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
dev
charleshall888
Development entry point — routes a request to the right workflow.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
discovery
charleshall888
Investigate a topic, then decompose it into backlog tickets grouped by epic.
SkillAFreshness: A★ 3Claude Code +9 - 0.606score
research
charleshall888
Parallel research orchestrator — dispatches 1–6 agents across independent angles, synthesizes into research.md.
SkillAFreshness: A★ 3Claude Code +9
Cite this page
refine skill. Agent Mart. https://agentmart.dev/skills/charleshall888-refine (last updated 2026-10-24)