repo-tour
Skill4 starsChanged 3 months agoMITView on GitHub
Use after graphify has already built graphify-out/graph.json for a project, to generate a dependency-ordered guided tour, an onboarding doc for new team members, or a diff-impact analysis before a commit/PR — reading the existing graph instead of re-scanning the repo. Trigger on "onboarding guide", "เอกสารสำหรับทีมใหม่", "guided tour ของโค้ด", "diff impact", "อธิบาย repo ให้ทีมใหม่", "what does this change affect", when graphify-out/graph.json already exists. If graphify-out/graph.json does NOT exist yet, tell the user to run graphify first — this skill does not scan or build a graph itself.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- thinkyou0714/github-flow-kittheir own version, first seen Oct 13, 2026
- Andersonnn7788/claude-code-101their own version, first seen Aug 30, 2026
- mae616/ai-templatetheir own version, first seen Aug 22, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- mae616/ai-template★ 9same name
- Andersonnn7788/claude-code-101★ 1same name
- thinkyou0714/github-flow-kit★ 0same 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 repository58 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/project-docs-prep/templates/AGENTS.mdAGENTS.md
- skills/project-docs-prep/templates/CLAUDE.mdCLAUDE.md
- deep-planningSkill
- deep-planningSkill
- deprecatedSkill
- architecture-reviewSkill
- code-reviewSkill
- diagnoseSkill
- tddSkill
- ui-design-engineeringSkill
- ux-ui-design-assessmentSkill
- in-progressSkill
- mcp-tool-managementSkill
- skill-managementSkill
- obsidian-canvasSkill
- obsidian-markdownSkill
- obsidian-notion-bridgeSkill
- obsidian-vault-workflowSkill
- boomz-preferencesSkill
- personal-agent-workflowSkill
- personalSkill
- automation-designSkill
- grill-with-docsSkill
- handoffSkill
- inbox-meeting-triageSkill
- markdown-reportSkill
- prompt-improvementSkill
- project-docs-prepSkill
- claim-citation-checkSkill
- daily-news-reportSkill
- github-skill-researchSkill
- project-discoverySkill
- research-methodologySkill
- seo-website-evaluatorSkill
- thaiSkill
- thai-addressSkill
- thai-customer-serviceSkill
- thai-date-formatSkill
- thai-festival-cardSkill
- thai-government-formSkill
- thai-id-validateSkill
- thai-invoiceSkill
- thai-pdpaSkill
- thai-resumeSkill
- thai-social-captionSkill
- thai-text-processingSkill
- thai-translateSkill
- backtest-expertSkill
- dividend-value-screenerSkill
- drawdown-circuit-breakerSkill
- exposure-coachSkill
- market-breadth-analyzerSkill
- portfolio-managerSkill
- position-sizerSkill
- tradingSkill
- trader-memory-coreSkill
- uptrend-analyzerSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Fast-path rule (check this first, every time): if graphify-out/graph.json does not exist, stop and tell the user to run graphify first …
- Distinctive: line 62
All four modules are pure stdlib (no dependencies) and independently self-testable:
- Very specific: line 69
All three CLIs write UTF-8 correctly to stdout by default, and to a file with --out <path> (parent directories are created automatically) — …
- Very specific: line 71
Run python3 <file>.py with no arguments for that module's self-test (uses synthetic in-memory graph data plus, for graph_adapter.py, a …
- Very specific: line 83
Never write any of the three into graphify-out/ itself — that directory is gitignored (it holds a machine-local absolute interpreter path …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.
Files
Repo Tour (reads an existing graphify graph)
Overview
graphify already builds a persistent knowledge graph (graphify-out/graph.json) from a codebase — that's the expensive, token-hungry part, and it's already solved. What's missing is three specific consumption modes: a dependency-ordered reading tour, an onboarding doc, and a diff-impact report. This skill supplies those three by reading the existing graph, not by re-scanning the repo — cheap, fast, and doesn't duplicate graphify's own extraction pipeline.
Fast-path rule (check this first, every time): if graphify-out/graph.json does not exist, stop and tell the user to run graphify first (/graphify or the graphify skill). Do not fall back to reading files manually — that's project-discovery's job, not this skill's.
Cost note before suggesting a graphify run: repo-tour itself makes no LLM calls and costs nothing to run. Building the graph in the first place is not free, though — graphify's semantic extraction pass (anything that isn't code: .md, .pdf, images) dispatches an LLM subagent per ~20-25 files, and a real test run in this repo (10 small Markdown files) used ~116k output tokens for that one pass. If no graph exists yet and the corpus is doc-heavy or large, say so explicitly and let the user decide whether to scope the graphify run narrower (a subdirectory, not the whole repo) rather than defaulting to a full-repo build. For a genuine one-off "explain this to me" question with no existing graph and no repeat use planned, project-discovery is usually cheaper — repo-tour's value shows up when a graph already exists (regular graphify use) or the same graph gets reused across multiple onboarding/diff-impact checks, not on a single cold-start question.
When to use
- สร้างเอกสาร onboarding สำหรับทีมใหม่ จากกราฟที่ graphify สร้างไว้แล้ว
- สร้างลำดับการอ่านโค้ดที่แนะนำ (จากโมดูลพื้นฐานไปหาโมดูลที่พึ่งพามัน)
- เช็คว่า diff/PR ปัจจุบันกระทบส่วนไหนของระบบบ้าง ก่อน commit
- Any request to "explain this repo to a new person" or "what breaks if I change X" when a graph already exists
When NOT to use
graphify-out/graph.jsondoesn't exist — rungraphifyfirst, this skill has nothing to read- The user wants the graph itself explored interactively (queries, path-finding between concepts) — that's
graphify query/graphify path/graphify explain, already built and more capable than anything here - A one-shot, no-graph project overview for a small/unfamiliar repo —
project-discoveryis lighter-weight for that
Core knowledge
Three sub-workflows, one shared graph loader (graph_adapter.py):
| Workflow | Script | Input | Output |
| Guided tour | build_tour.py | graph.json | Dependency-ordered node list (foundational modules first) |
| Onboarding doc | build_onboarding.py | graph.json + project name | Markdown doc: hub modules + reading order |
| Diff impact | diff_impact.py | graph.json + changed file paths | Per-file impact report, grouped by hop distance |
Guided tour ordering: topological sort over imports/depends_on/requires edges. An edge source → target means "source depends on target," so the tour visits target first. Real codebases have circular imports — nodes caught in a cycle are appended afterward and explicitly flagged cycle_fallback, never silently mixed in as if they were cleanly ordered.
Onboarding doc: ranks "hub" modules by total edge degree (most-connected = worth understanding early), computed directly from the edge list — does not depend on graphify's god_nodes/community analysis files, which graphify deletes as intermediate scratch after each run and are not guaranteed to exist.
Diff impact: for each changed file, walks the graph backward (finds what depends on the changed file, not what the changed file depends on) up to a hop limit, grouped by distance. Get the changed-file list with:
git diff --name-only # unstaged changes
git diff --name-only --cached # staged changes
git diff --name-only HEAD~1 # vs. the previous commitThen run: python3 diff_impact.py graphify-out/graph.json <file1> <file2> ...
This is a read-only report over the graph — it does not run tests or execute anything. It narrows where to look, it does not replace actually running the test suite.
Common mistakes
- Running this skill without checking
graphify-out/graph.jsonexists first — there's nothing to read, and re-scanning the repo manually here duplicates graphify's job. - Treating a
cycle_fallbacknode in the tour output as if it were cleanly topologically ordered — it wasn't; the graph has a circular dependency there. - Assuming diff-impact's "no dependents found" means the change is definitely safe — it means the graph shows no dependents; a genuinely new/untracked relationship (or a stale graph) can hide real impact. Suggest re-running graphify if the graph is old relative to the diff.
- Treating the impact report as a substitute for running tests — it's a "where to look" map, not a correctness guarantee.
Code
All four modules are pure stdlib (no dependencies) and independently self-testable:
graph_adapter.py— schema-tolerant loader (load_graph,normalize_node,normalize_edge,build_adjacency), plus the shared--outflag helpers (parse_out_flag,write_or_print) all three CLIs below usebuild_tour.py—topological_tour(nodes, edges); CLI:python3 build_tour.py graphify-out/graph.json [--out tour.md]build_onboarding.py—render_onboarding_doc(project_name, nodes, edges); CLI:python3 build_onboarding.py graphify-out/graph.json "Project Name" [--out onboarding.md]diff_impact.py—impact_of_changed_files(nodes, edges, changed_paths); CLI:python3 diff_impact.py graphify-out/graph.json <changed files...> [--out impact.md]
All three CLIs write UTF-8 correctly to stdout by default, and to a file with --out <path> (parent directories are created automatically) — equivalent to > file.md but without depending on the shell's own redirect encoding behavior (see the earlier em-dash/Windows finding above).
Run python3 <file>.py with no arguments for that module's self-test (uses synthetic in-memory graph data plus, for graph_adapter.py, a fixture in the real graphify shape — see below).
Where to put each output — they don't all belong in the same place
The three outputs have different lifecycles; treat them differently, not identically:
| Output | Persist it? | Where | Why |
| Onboarding doc | Yes, commit it | docs/ONBOARDING.md or repo root | Meant to be read by multiple people, repeatedly, over time — a durable artifact |
| Guided tour | Optional | docs/ alongside the onboarding doc | Nice to keep as a reference, less critical than the onboarding doc |
| Diff impact | No — don't commit as a standing file | Paste into the PR description or chat directly | It describes one specific uncommitted diff; the moment that diff changes or merges, a saved file describing it is stale and misleading to a later reader who doesn't know it's from an old snapshot |
Never write any of the three into graphify-out/ itself — that directory is gitignored (it holds a machine-local absolute interpreter path and regenerable cache), so anything meant to be shared or kept must go somewhere your VCS actually tracks.
Verified against a real graph (not just guessed)
The schema in graph_adapter.py was checked against graphify's actual installed source (graphify/export.py's to_json()) and confirmed by running the full pipeline on skills/trading/ end to end. Two real bugs were found and fixed this way — worth knowing since they show what "looks right but silently isn't" looks like for this kind of tool:
- The top-level edge list key is
"links", not"edges"— graphify callsnetworkx.readwrite.json_graph.node_link_data(G, edges="links"), which renames the key. A naivegraph["edges"]on a realgraph.jsonsilently returns nothing (no error, just an empty list) rather than failing loudly.graph_adapter.pynow checks"links"first, falling back to"edges". - Node type lives in
"file_type"(code/document/paper/image/rationale/concept), not"type". Node file path lives in"source_file", not"filePath"/"path". Both are now the first-checked key. - The real AST edge type for a Python import is
"imports_from", not"imports".build_tour.py'sDEPENDENCY_EDGE_TYPESanddiff_impact.py'sIMPACT_EDGE_TYPESboth check for it now — omitting it silently produced correct-looking but incomplete output (no crash, just a tour/impact report missing real dependency edges). find_node_by_path's tolerant suffix matching was too loose for short generic filenames. graphify relativizes a root-level skill's own file to a bare"SKILL.md"with no directory — the original suffix check (file_path.endswith(node_path)) then matched any longer path ending in/SKILL.md, e.g. querying"drawdown-circuit-breaker/SKILL.md"wrongly matched the router's bare"SKILL.md"node instead of (or in addition to) the actual target. Fixed by requiring the shorter of the two paths to contain at least one/before suffix-matching is allowed — a bare filename now only matches by exact equality.
Known limitations
- Onboarding "hub" ranking (edge degree) is a simple proxy for importance, not graphify's own god-node algorithm (which may use different centrality measures) — the numbers won't necessarily match
GRAPH_REPORT.md's own god-node list if that file is still around. - Diff-impact only follows structural edge types (
imports,imports_from,depends_on,requires,calls) — it will not surface a change's impact through doc-level relations (references,cites,conceptually_related_to) even though those exist in the same graph. A change to aSKILL.mdthat other skills only reference in prose, not import in code, will correctly report "no dependents" — that's accurate for code-impact, not a bug, but don't read it as "nothing in the project mentions this file." imports_fromedges can point at a synthetic node for an external/stdlib symbol (e.g.Decimal), not just another project file — those show up in the guided tour as an ordering step with no further meaning; harmless, but don't be surprised to see a stdlib name in the reading order.- This was verified against one real run (
skills/trading/, a small, mostly-Python-with-docs corpus) on one graphify version — a different corpus shape (e.g. TypeScript, or a run with--mode deep) or a future graphify release could still surface a field this skill hasn't seen yet.graph_adapter.py's defensive multi-key-variant lookups remain the safety net for that.
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.

