walkthrough
Skill0 starsChanged 40 days agoMITView on GitHub
Work through a list point by point — open questions, review findings, plan sections, options — each with a brief, a recommendation and a picker, tracked in a decision log. Also runs guided tours of a large document or repo. Not for a plain "explain X" question. Use when the user says "go through these one by one", "point by point", "walk me through the review/plan", "guide me through this document/repo step by step", "give me a tour of", "continue the walkthrough", or accepts such an offer.
34 other projects use a version of this.
Adopted by
Most recently taken up by
- moisesrj97/skillstheir own version, first seen Nov 16, 2026
- sandoche/agentic-code-unpackedtheir own version, first seen Nov 16, 2026
- iangregsondev/agent-skillstheir own version, first seen Nov 16, 2026
- udecode/dotaitheir own version, first seen Nov 13, 2026
- pche-broken-artist/forge-of-thoughttheir own version, first seen Nov 10, 2026
- taecontrol/skillstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use chips, redo and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
udecodeSkillchips
Run the project's suggested-task ledger, the chips. /chips alone brings this session's open tasks up to date: it applies what the ledger helper says, settles stale and split tasks, cards what can start and reports the board, and /chips all also takes the ship-list tasks and the ones no live session owns, and merges duplicate tasks across the board. Use for /chips, /chips offer, start or drop, \"what chips are open\", \"suggest a task for X\", \"update the chips\", \"clean up the chips\", \"which chips are stale\", \"is this chip still worth doing\", \"merge duplicate chips\", at the last hand-back of a run that offered a task, and whenever pstack would file a tracked issue, a follow-up, a reflect Backlog item or an out-of-band PR.
Also in 3 projects that use this
Scylla23Skillredo
Re-run the last routed task one model tier up (haiku → sonnet → opus → fable). User-invoked escalation when the previous result missed. Pass --dry-run to preview without running.
Also in 3 projects that use this
Aetheria-Labs1Skillplan-page
Plan one storefront page from strategy through final copy, section order, proof, offers, template direction, responsive layout, and production asset requirements. Stops at an approved execution-ready specification.
Also in 3 projects that use this
SocialKit-APISkillvideo-transcripts
Extract transcripts from social videos through SocialKit across YouTube, TikTok, Instagram, Facebook, X/Twitter, LinkedIn, and direct video file URLs. Use when the user asks to transcribe, get captions or subtitles, pull the spoken text, or read what a video says, including one video or many at once. Returns full text plus timestamped segments. Not for posting, private accounts, or unrelated website scraping.
Also in 3 projects that use this
MicrosoftSkillcleanup
Also in 3 projects that use this
BlockSkilltest-audit
Invoke whenever writing, changing, reviewing, or sweeping tests in buzz-app. Authoring gate for new tests plus audit workflow for low-value, implementation-coupled, or duplicative tests and the test-only production seams they demand.
Also in 3 projects that use this
Other versions34 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 34
- udecode/plate★ 17ksame name
- udecode/dotai★ 1.2ksame name
- udecode/better-convex★ 452same name
- Mathews-Tom/armory★ 340same name
- smallnest/goal-workflow★ 283same name
- andrew-yangy/gru-ai★ 153same name
- oborchers/proqi★ 45same name
- huytd/skills★ 19same name
- wycats/ai-plugin★ 19same name
- wycats/vscode-ai-plugin★ 19same name
24 more not listed
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.
The file
Files
/walkthrough — point-by-point review and guided tours
Arguments: $ARGUMENTS
Hard rules
- One point at a time, each with a progress header — the user never faces a wall of questions. Only clear-cut ✓ points may be batched (≤4, a 1–2 line brief each, header
Points 4–7/10). - No point is dropped. The walkthrough ends only when every agenda item has an outcome, or the user explicitly leaves it open (recorded as open).
- The log is visible: state its path at the start and in every "recorded" line.
- Re-check facts against the real code and docs before presenting each point; correct earlier claims openly ("Correction: …"), including your own earlier review.
- Never change a
Status:line or a plan gate line, and follow${CLAUDE_SKILL_DIR}/../impl-plan/reference/doc-model.mdwhenever the source is a plan or milestone doc (where to record: §3.5) — plusdoc-authoring.mdnext to it when a decision rewrites plan text. - Never smuggle in content: apply only what was decided; report anything added or dropped that wasn't. Never copy credentials or tokens into a log.
- Ledger only via
/followup. Never commit — suggest the message (or, when invoked from another skill's flow, leave the commit to that flow). - Edits outside the decisions section wait for confirmation (checkpoint or end). Hands-off files are never edited — flag them instead: anything outside the repo root, submodules, paths the project CLAUDE.md marks hands-off or as frozen specs, doc-sync's declared no-touch docs, generated files.
- Plan mode active → write nothing (no log file, no doc edits, no
/followup): keep the log in the conversation and, at the end, list the log, decisions and follow-ups to write once plan mode is off. - Tour mode is information-only — follow
${CLAUDE_SKILL_DIR}/reference/tour.md, which replaces §2–§4. It decides nothing and edits no source; it writes only on request (a notes file, or/followupentries for parked decisions at the end). All hard rules still apply, except that there is no log unless notes are saved — then "the log is visible" applies to the notes file.
1. Source
| Argument | List to walk through |
| (none) | The list in the most recent substantive response or relayed agent report — questions, findings, plan points, options. If the last response offered a tour, start that tour on the offered target. No list (or two, or fewer than 3 items) → ask what to walk through, offering candidates: the last agent report, open D<n> in the active milestone doc, open Kind: decision ledger entries, a file path, a tour of a file or this repo. |
<file> | Its open items: undecided design questions (D<n> / Q<n>), review findings, proposed changes, plan sections, ledger entries with Kind: decision. A file or section with no open items → say so (name what's already decided) and offer a tour — never start one unasked. A directory → ask once: the open decisions across its docs, or a tour. |
tour [<file | dir | . | topic>] [goal] | A guided, information-only tour of a document, a repo or directory, or a topic from this conversation — ${CLAUDE_SKILL_DIR}/reference/tour.md replaces §2–§4 below (argument rules there). |
<file>#<section> | Only that section (use the doc's own heading — e.g. its design-questions section). |
resume [<log>] | Unfinished logs in the walkthrough log dir: one → continue it; several → list them (date · source · open count) and ask; offer to mark stale ones abandoned. If a <file> argument names a different source than a log, start a new walkthrough. No log, but an unfinished chat-only walkthrough or tour in this conversation → continue it; none → say so and offer to start one. |
--only 2,5,7 | Restrict to those agenda items (recorded as the log's filter). |
2. Agenda
- Extract the items; merge duplicates, split compound ones. Keep the source's grouping and labels (Fix 2, D3, Finding 7) as stable IDs beside the running number — never renumber; a merged item keeps its label and shows
→ covered by #3. - Order: groups as in the source; within a group, decisions that constrain others first, then the most important.
- Mode (say which): decide (questions/options) · review (findings or proposed changes: accept / modify / reject) · tour (explain each stop; nothing to decide —
reference/tour.md). A decide/review list can include explain-only points: they are ✓ with the recommendation "explain", use the tour picker and the outcomeunderstood, and under a bulk option are shown as a batched 1–2 line brief each — never recordedunderstoodunseen. A list with both kinds is logged asMode: mixed. - Show the agenda as a compact table:
# · ID · point · recommendation · ⚖ judgment call / ✓ clear-cut. - Ask once how to proceed (AskUserQuestion): One by one (Recommended) · Only the judgment calls — accept the rest · Pick points (then type the numbers) · Accept all recommendations. Bulk-accepted points are still recorded; any that carries a compromise still becomes a follow-up.
Create the log from
${CLAUDE_SKILL_DIR}/templates/log.md, with a stub for every point (1–3 lines of substance, plus any proposed text verbatim) so the walkthrough survives a new session even when its source was a chat response. Where:- the reviews path declared in the project CLAUDE.md's Documentation Workflow section, else
<repo root>/docs/reviews/(repo root fromgit rev-parse --show-toplevel); fileYYYY-MM-DD-<slug>.md, adding-2,-3if taken; - ask first ("create
docs/reviews/for the log, or keep it in chat?") when the project has nodocs/dir, when the folder isn't a git repo, or when the source is a conversation that isn't about this project's code or docs (personal or unrelated content defaults to chat only); - tours default to chat only (notes on request —
reference/tour.md).
Print the path (or "log kept in this conversation — it won't survive a new session").
- the reviews path declared in the project CLAUDE.md's Documentation Workflow section, else
3. Each point
- Header:
**Point 3/10 · Fix 4 — <title>**(⚖ if it's a judgment call) ·<n> left. Brief (~150–300 words for a fresh point; building out a user's idea may run longer):
- the situation in 1–3 sentences, with
file:linereferences; - options A / B (/ C), each with its tradeoff or cost;
**Recommend A** — <why>; interactions with earlier decisions;- when the point changes a document: the exact proposed text in a blockquote, where it goes ("replaces X in §Y"), and side effects elsewhere — approving means accepting that edit;
- sub-decisions numbered with a stated lean ("1. … — I lean toward …");
- external facts marked [V] (verified, with source) or [U] (unverified or inferred).
Review mode: the finding, the proposed change, its risk. Explain-only points: the explanation, as a tour stop.
- the situation in 1–3 sentences, with
Picker — AskUserQuestion;
header= the point's ID (≤12 chars, e.g.D3,Fix 4); short labels, tradeoffs in each option's description; "Other" is automatic and always works for refining, own answers, questions and navigation:- decide:
<A> (Recommended)· up to two more alternatives ·Tell me more— with only one alternative, addDefer as follow-up; further alternatives stay in the brief and are chosen via Other; - review:
Accept (Recommended)·Reject·Tell me more·Defer as follow-up("accept with changes" goes in Other); - explain-only / tour:
Got it — next·Expand·Skip — review later(questions go in Other and are answered on the spot).
Sub-decisions: the main question plus at most three sub-questions in the same call, each with 2–4 options; Tell me more / Defer only on the main one. Free-text replies are fine: "approve", "your call", "leave it to you" = decided as recommended. No AskUserQuestion available (headless runs) → the same choices as a numbered plain-text list.
- decide:
- Handle the reply per
${CLAUDE_SKILL_DIR}/reference/replies.md. Record immediately:
- the log: the point's status, outcome, rationale, rejected options, follow-ups, queued edits;
- the source's decisions section, by source:
- milestone doc,
proposed→ rewrite theD<n>bullet inoverview.mdto its decided form (<chosen>. Rejected: <alt> — <why>. (<date>, user); a point the user wants to revisit also gets a/followuplinked from it); a new design question gets the next freeD<n>bullet.approved→ same, then tell the user it's a substantive change:/milestone refresh <id>re-approves (Status returns toproposedthere, not here).in progress/landed/superseded, or a project that freezes design docs → log only, and suggest/milestone refresh <id>or a dated As-built note; - plan
draft→ write the decision into §Decisions as Chosen vs Rejected and resolve its[VERIFY]tag / register row (Answer + Basis) or re-own it.approved→ the same is a reviewed reversal: date it in §Decisions, name it in the suggested commit message, and warn which milestone docs it makes stale; - ledger → a
/followupUpdate bullet on the entry ("confirmed / changed in walkthrough<log>#N"); status flips stay with doc-sync; - anything else (a chat response, a review report) → log only;
- milestone doc,
- other implied edits (review fixes, aligning other docs) are queued — later points may change them.
Then one line:
Recorded in <log> (+ <doc>) · follow-ups: FU-NNN · Next: Point 4/10 · <ID> — <title>.- Pace: after three straight accepts in a row, or at a group boundary, offer "Accept the rest of
<group>as recommended?" — once per group; judgment calls are still asked. At group boundaries also offer a checkpoint apply: list the approvals since the last apply, apply on yes.
Navigation (any time, via Other or free text): next · back to N · jump to N · skip (returns at the end) · accept the rest · only judgment calls left · summary so far · apply what we have · pause (the log keeps the state; /walkthrough resume continues).
On resume: re-read the source; if it changed since the log's Source at, re-verify the open points and queued edits (and, for a milestone doc, that its Status still allows writing) before continuing at the first open point with a one-line recap.
4. End
- Completeness check: list every point without an outcome (skipped, open); ask: go through them now / defer each as a follow-up / leave open (recorded as open). Then say it plainly: "Every point from
<source>now has an outcome" — or name what's left. - Summary table:
# · ID · point · outcome · follow-ups+ the log path. Outcomes: decided · deferred (FU) · future direction · handed off · covered by #N · understood · skipped · open. - Queued edits since the last checkpoint, listed by file; ask once "Apply these?"; apply on yes. Then report anything added or dropped that wasn't explicitly decided, and remove it unless the user keeps it. Set the log's Status to
done. - Invoked from another skill's flow (
/milestone refreshapproval,/impl-planhand-off) → stop here and continue that flow's next step in the same turn; the log and ledger entries go into its commit. Otherwise — what's next: for a milestone source, "Status stays<status>;/milestone refresh <id>(or/implement <id>) approves it"; recommend the next step, offer a ready-to-paste prompt if it belongs in a new session, and suggest a commit message.
Recording each point edits files — in default permission mode, suggest accepting edits for the session up front so the walkthrough isn't interrupted by prompts.
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
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.

