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/paulnsorensen-cure
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: agent-building
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/cure
What's inside
- Skill name
cure- Description
- Finish what /age started. Loads both sidecars (fixes + suggestions), renders a unified stake table, gates on user approval, routes each approved item to the right handler (/cleanup or a cook sub-agent), then re-ages the touched paths up to a hard 3-turn cap.
- License
- MIT
- Metadata
- owner: cheese-flow · category: cleanup
- Instructions
- 134 lines in SKILL.md · view on GitHub
Agent Skills spec issues
- allowed-tools: must be a space-separated string
Files (4)
- SKILL.md
- apply-router.md · references/
- re-age.md · references/
- sources.md · references/
SKILL.md (first lines)
# Cure — Finish what `/age` started
`/cure` is the *single* post-`/age` finisher. Where `/age` reviews and
emits sidecars, `/cure` consumes them, gates on the user, applies items
through the right handler, and re-ages the touched paths. In
cheesemaking, curing follows aging — the verb is exact.
The loop is intentional and bounded:
```
load → user gate → apply → re-age → (turn)
```
Nothing applies without explicit user approval. The re-age verify step
has a hard 3-turn cap per invocation.
## Arguments
```
/cure <slug>
```
`slug` is the same slug emitted by `/age`. `/cure` reads
`.cheese/age/<slug>.fixes.json` (required) and
`.cheese/age/<slug>.suggestions.json` (optional). See
`references/sources.md` for the resolution table and the
missing-sidecar error contract.
All harnesses share the project-root `.cheese/` runtime directory.
## Phase 1 — Load
Load `.cheese/age/<slug>.fixes.json` (required) and the optional
companion `.cheese/age/<slug>.suggestions.json`. Merge their items into
a single list. If the fixes file is missing, fail fast with the
missing-sidecar error documented in `references/sources.md`.
Fix items and suggestion items have different required keys:
- **Fix items** (`fixes.json`): `id`, `dimension`, `file`, `anchor`,
`content`, `rationale`, `category`. All mechanically-applicable via
`tilth_edit`.
- **Suggestion items** (`suggestions.json`): `id`, `dimension`, `file`,
`outline_ref`, `narrative`, `agent_brief_for_cook`. No anchor or content;
not mechanically-applicable.
## Phase 2 — User gate
Render the merged items as a stake-grouped table (high → medium → low),
sorted by file within each group:
```
| id | stake | category | dim | location | summary |
```
`stake` is not a stored field — derive it from `dimension` using the
fixed map: high (correctness, security, encapsulation, spec), medium
(complexity, deslop, assertions, nih), advisory (precedent). `category`
shows the sub-type for fix items (e.g. `deslop.swallowed_catch`) or
`suggestion` for suggestion items.
Items are already pre-classified by `/age`; `/cure` does not
re-classify. The default selection is **empty** — nothing applies on a
bare return. Recognized verbs:
```
1,3,5 (specific ids)
all-high (every high-stake item)
none (default; exit cleanly)
skip N (drop item N from the change-order)
```
No item executes without an explicit selection. `/cure` always waits for
the user to pick ids; the orchestrator does not run on the user's behalf
and does not chain from `/age`.
## Phase 3 — Apply
Each approved item dispatches by **sidecar of origin** through the apply
router (see `references/apply-router.md`): `fixes.json` items route as
`edit`; `suggestions.json` items route as `suggestion`. The `category`
sub-type (e.g. `deslop.swallowed_catch`) is informational only.
| routing type | handler |
|---|---|
| `edit` | `/cleanup <slug>` (single-item synthesized sidecar) |
| `suggestion` | spawn one cook sub-agent per item with `agent_brief_for_cook` |
Cross-slice calls go through public skill entries only (`/cleanup`,
`/age`, `/cook`). No reaching into sibling internals.
For each applied item, record `touched_paths += item.file`.
## Phase 4 — Re-age
If any paths were touched, run internally:
```
/age --scope <touched_paths>
```
Diff the resulting `.cheese/age/<slug>.{fixes,suggestions}.json` against
the prior items and emit only **new or changed** items as the next
iteration. If empty, the loop exits cleanly.
If non-empty, ask the user whether to continue into the next turn. The
loop has a hard **3-turn cap** per invocation (`turn <= 3`, 1-indexed).
After turn 3, force-exit and surface remaining items as a one-line
"next session" hand-off the user can run as `/cure <slug>` again. See
`references/re-age.md` for the full diff semantics, the turn log, and
the cap rationale.
## Rules
- Default selection is empty. The user gate is the only path to apply.
- `/cure` is invoked manually after `/age` prints the hand-off; the
orchestrator never chains it on the user's behalf.
- File I/O via `cheez-read` / `cheez-search` / `cheez-write`. No host
`Read` / `Grep` / `Edit`. Sidecar JSON loads through `cheez-read`;
direct edits use `cheez-write`. Hash-anchored fixes are owned byFrom paulnsorensen/cheese-flow, MIT.
More from paulnsorensen/cheese-flow
- 0.498score
nih-audit
paulnsorensen
Scan a codebase for custom code that duplicates what open-source libraries already do, then recommend which libraries to adopt.
SkillCFreshness: C★ 1Claude Code +9 - 0.497score
cheez-search
paulnsorensen
AST-aware code search using tilth MCP.
SkillCFreshness: C★ 1Claude Code +9 - 0.495score
cleanup
paulnsorensen
Mechanical apply of /age's hash-anchored sidecar fixes via tilth_edit.
SkillCFreshness: C★ 1Claude Code +9 - 0.495score
age
paulnsorensen
Staff Engineer code review orchestrator.
SkillCFreshness: C★ 1Claude Code +9 - 0.494score
mold
paulnsorensen
Iterative thinking amplifier for fuzzy ideas.
SkillCFreshness: C★ 1Claude Code +9 - 0.490score
milknado-execute
paulnsorensen
Use the Milknado MCP tools to pick and execute the next ready node from the Mikado graph, apply file changes, and run quality gates.
SkillCFreshness: C★ 1Claude Code +9 - 0.490score
research
paulnsorensen
Multi-source research orchestrator that routes questions across Context7 library docs, Tavily web research, codebase analysis, and GitHub examples, then…
SkillCFreshness: C★ 1Claude Code +9 - 0.437score
merge-resolve
paulnsorensen
Resolve merge conflicts, rebase conflicts, and cherry-pick failures using mergiraf (AST-aware structural merge), git rerere, and kdiff3.
SkillDFreshness: D★ 1Claude Code +9 - 0.432score
gh
paulnsorensen
Complete GitHub tasks using the gh CLI.
SkillDFreshness: D★ 1Claude Code +9 - 0.432score
cheez-read
paulnsorensen
Smart code reading using tilth MCP.
SkillDFreshness: D★ 1Claude Code +9 - 0.432score
cheez-write
paulnsorensen
Hash-anchored file editing using tilth MCP.
SkillDFreshness: D★ 1Claude Code +9 - 0.425score
milknado-plan
paulnsorensen
Use the Milknado MCP tools to decompose a goal into a Mikado-style change graph and produce a token-budgeted batch plan.
SkillDFreshness: D★ 1Claude Code +9
Signals
- Stars
- 1
- Forks
- 1
- Open issues
- 25
- Last commit (this folder)
- May 3, 2026 (6 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/paulnsorensen/cheese-flow
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from paulnsorensen
- 0.603score
build-optimize
paulnsorensen
Optimize local build, test, and check commands with measured wall-time evidence.
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
chezmoi
paulnsorensen
Guides dotfile management with chezmoi when the user says "set up dotfiles", "manage my dotfiles", "bootstrap a new machine", "chezmoi", "encrypt my SSH key…
SkillAFreshness: A★ 2Claude Code +9 - 0.603score
bin-doctor
paulnsorensen
Audit and repair installed binaries so each tool has one install path, a clean cache, and a current pin, across mise, Homebrew, cargo, uv, npm, pnpm, bun…
SkillAFreshness: A★ 2Claude Code +9 - 0.600score
ci-optimize
paulnsorensen
Optimize GitHub Actions workflow wait time with measured evidence.
SkillAFreshness: A★ 2Claude Code +9 - 0.598score
cheese-board-loop
paulnsorensen
Runs the cheese-board workflow and asks the user each question it raises.
SkillAFreshness: A★ 2Claude Code +9 - 0.592score
skillz
paulnsorensen
Add, improve, audit, wedge, or autoimprove a skill.
SkillAFreshness: A★ 1Claude Code +9
Cite this page
cure skill. Agent Mart. https://agentmart.dev/skills/paulnsorensen-cure (last updated 2026-10-24)