agentleFS
victor-hac-workvictor-hac-work/ai-unknowns-resolver/skills/impl/SKILL.md

impl

Skill0 starsChanged 4 months agoMITView on GitHub

Implement a task planned with unknowns:pre. First generates an implementation PREVIEW (diagram + pseudocode + file-change map as HTML) the user must critique and approve, then implements with the deviation-logging convention (conservative choice, log it, keep going — no stopping to ask), and appends newly discovered unknowns to the ledger. Use in a FRESH session after /unknowns:pre finishes, or when the user says "/unknowns:impl" or "implement the unknowns plan".

3 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions3 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 3

    1. Noizefield/audio-plugin-coder★ 334same name
    2. devjiwonchoi/skills★ 2same name
    3. Daeguk-Sun/dcNess★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No matter how much planning was done, unknown unknowns lurk in implementation.

    line 8
  2. Very specific:

    No argument: if exactly one directory exists under docs/unknowns-resolver/, use it.

    line 13
  3. Very specific:

    Read, in order: ledger.md, 06-plan.md, 04-prototype/reactions.md (if present), 05-reference.md (if present).

    line 18
  4. Distinctive:

    Before touching any code, show the user what the implementation will look like:

    line 22
  5. Very specific:

    Write 07-preview.md and generate 07-preview.html per ${CLAUDE_PLUGIN_ROOT}/skills/pre/references/design-system.md (§ 07-preview.html).

    line 24
  6. Very specific:

    Run the plan's ## Verification commands.

    line 75
  7. Very specific:

    Update ledger (Route status → impl complete), regenerate index.html (per ${CLAUDE_PLUGIN_ROOT}/skills/pre/references/design-system.md — …

    line 79

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

unknowns:impl — preview, approve, then implement with a deviation log

No matter how much planning was done, unknown unknowns lurk in implementation. This phase shows the user WHAT is about to be built (cheap to change), gets approval, then implements the plan AND captures what the plan missed, so the next attempt starts smarter.

Step 1 — Locate state

  • Argument given (/unknowns:impl docs/unknowns-resolver/<slug>/) → use it.
  • No argument: if exactly one directory exists under docs/unknowns-resolver/, use it; otherwise list them and ask which (one question).
  • Directory or 06-plan.md missing → STOP and print: "No plan found. Run /unknowns:pre <task> first." (Do not improvise a plan.)

Step 2 — Load context

Read, in order: ledger.md, 06-plan.md, 04-prototype/reactions.md (if present), 05-reference.md (if present). Restate in ≤6 bullets: what's being built, the key decisions (by ledger #), and where the improvise points are.

Step 3 — Implementation preview (APPROVAL GATE — no code before approval)

Before touching any code, show the user what the implementation will look like:

  1. Write 07-preview.md and generate 07-preview.html per ${CLAUDE_PLUGIN_ROOT}/skills/pre/references/design-system.md (§ 07-preview.html). Contents, all derived from 06-plan.md + a read of the ACTUAL target code:
    • Architecture diagram — components touched and how they connect after the change (new pieces highlighted).
    • File-change map — table: file → create/modify → purpose → plan section #.
    • Pseudocode — per major component, short and language-neutral, linked to the plan section it implements.
    • Data flow — the main path end-to-end (request → … → response, or input → output), with improvise points marked amber where they sit.
  2. Open 07-preview.html for the user (open <path>), then invite critique IN CHAT: "React to the preview — wrong shape, missing case, better split? I revise until you approve."
  3. On each critique: revise 07-preview.md, regenerate the HTML, summarize what changed in one line. Repeat.
  4. Gate: proceed to Step 4 only on explicit approval (e.g. "approved", "go", "implement it"). Record at the top of 07-preview.md: Approved: <short user wording> (revision <n>). If a critique changes a decision (not just shape), update the ledger row and 06-plan.md section it touches before re-rendering.
# Implementation preview: <task>

Approved: — (pending)   ·   Revision: 1

## Architecture
<diagram description mirrored from the HTML, in text>

## File-change map
| File | Action | Purpose | Plan § |

## Pseudocode
### <component>   (plan § <n>)
...

## Data flow
1. ...

Step 4 — Implement

Work the plan top to bottom (data model → interfaces → user-facing → mechanical). Follow the target project's own conventions and verification commands (its CLAUDE.md rules win on style/test matters). The approved preview is the shape contract — deviate from it only via the deviation rule below.

Deviation rule — when an edge case forces a departure from the plan or the approved preview (or an improvise point fires):

  1. Pick the CONSERVATIVE option (least surface area, most reversible).
  2. Log it in 08-notes.md under ## Deviations — then KEEP GOING. Do not stop to ask.
# Implementation notes: <task>

## Deviations
### D1: <what forced it>   (plan section <n>, ledger #<id> if improvise point)
- Chose: <conservative option> — Rejected: <alternative> — Why: <1 line>

## New unknowns discovered
- <question> → added to ledger as #<id>
  1. Brand-new unknowns (not deviations, genuinely new questions) → append to ledger: Quadrant UU, Found in 08-notes, Status open.

Step 5 — Verify

Run the plan's ## Verification commands. Report results honestly — failures are failures.

Step 6 — Exit handoff (MANDATORY)

Update ledger (Route status → impl complete), regenerate index.html (per ${CLAUDE_PLUGIN_ROOT}/skills/pre/references/design-system.md — deviations panel now populated, next-action banner → post). Print:

Implementation complete: {sections done}/{total} plan sections, {d} deviations logged, {u} new unknowns. Preview: 07-preview.html (approved rev {n}) · Notes: 08-notes.md · Dashboard: index.html

Next phase: /unknowns:post docs/unknowns-resolver/<slug>/ — builds the buy-in pitch, then quizzes you. Merge only after a perfect quiz.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.