godmode-codegraph
Skill6 starsChanged 3 months agoApache-2.0View on GitHub
Query a local, on-demand code dependency graph to answer who calls, who imports, and what blast radius a change carries, before spending an unnecessary full read on a large module, refreshing the saved graph from the archive first when it looks stale. Use before an edit whose impact or required retests are not yet named. Not for a routine read with no graph question involved, or for chasing an already-failing check.
In this repository22 other files it gives its agents.
- llms.txtllms.txt
- godmode-budgetSkill
- godmode-changelogSkill
- godmode-ciSkill
- godmode-code-of-lawSkill
- godmode-continuitySkill
- godmode-evidenceSkill
- godmode-governanceSkill
- godmode-host-syncSkill
- godmode-impact-gateSkill
- godmode-investigationSkill
- godmode-loop-wardenSkill
- godmode-memory-gardenerSkill
- godmode-repairSkill
- godmode-replicateSkill
- godmode-researchSkill
- godmode-second-lookSkill
- godmode-skill-evalSkill
- godmode-skill-forgeSkill
- godmodeSkill
- godmode-spec-lifecycleSkill
- godmode-triageSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
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
- Distinctive: line 10
Before a change lands, its blast radius - the obligations, modules, and required retests that depend on what is about to change - is …
- Very specific: line 77
Never trust a graph's impact or must_retest result.
- Very specific: line 79
Never run atlas graph rebuild merely to erase a failing verify: that destroys the saved snapshot's only record that it had drifted, so fix …
- Distinctive: line 82
Never read a large file in full before checking the graph: query it first for who calls it and what it would break.
- Very specific: line 84
Never guess at a node id after unknown_node: true - rebuild or confirm the exact prefixed form instead.
- Very specific: line 86
Never skip atlas closure before a change described as low-risk.
- Distinctive: line 96
If an assertion cannot be proved, report the unmet assertion and the next safe action rather than the outcome it was meant to prove.
≈ “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.
Files
Godmode Codegraph
Outcome
Before a change lands, its blast radius - the obligations, modules, and required retests that depend on what is about to change - is answered from a saved, on-demand graph derived from this project's own archive, rather than by opening every file that might be affected.
Use
- Who depends on this module, obligation, or file before it is edited.
- Which tests must be retested for a set of changed paths.
- The saved graph looks stale (a rebuild is needed) or its snapshot needs proving against the archive it was derived from.
- A large file's dependents should be read from the compressed graph before a raw full-file read is spent on it.
Do Not Use
- Reading one small file's own contents when nothing needs its dependents - read the file directly.
- Investigating why one specific test already fails - that is
godmode-investigation. - Rebuilding the loop-record signature history for a stalled retry - that is
atlas loop, a different subcommand with a different evidence shape.
Deterministic Execution Flow
- Prove the saved snapshot before trusting it.
godmode atlas graph verify- rebuilds the graph fresh in memory (without saving it) and compares its hash against the last saved snapshot; reportsverified: falsewith areasonnaming why -no graph snapshot found; run atlas graph rebuildon first use,graph snapshot is stale; run atlas graph rebuildonce the archive has moved on - and leaves the saved snapshot untouched either way. Needs an initialized archive (godmode initfirst on a project that has none). - Rebuild only when step 1 says to.
godmode atlas graph rebuild- derives nodes and edges fresh from the archive's own records (obligations and theirblocked_byedges, sprint items,retested_byattestations, supersession, and citations) and overwrites the saved snapshot. Run this only afterverifyreports no snapshot or stale - never to make a failingverifypass, since overwriting the snapshot destroys the only record that it had drifted. - Ask the graph, not the files.
godmode atlas graph query <node> [--depth N]- node ids look likeobligation:<subject>,module:<dotted.name>, orfile:<path>; loads the snapshot step 1 proved rather than rebuilding. Returnsimpact(what transitively depends on the node over currentdepends_onedges) andmust_retest(attestations reachable overretested_byedges).- Fallback: an id the graph has never seen returns
unknown_node: trueand refuses (exit 1) rather than guessing - rebuild first, or confirm the id's exact form (obligation:/module:/file:prefix).
- Fallback: an id the graph has never seen returns
- Name what a working-tree change touches.
godmode atlas closure [--changed PATH...] [--depth N]- dependents of what changed that were not themselves changed; defaults to the current working tree when no paths are given. Refuses (exit 1,graph unverified: run atlas graph rebuild) if the saved snapshot fails its own verify, re-checking step 1's proof before trusting anything. Exits 0 even when nothing changed (verdict: nothing-changed). - Run only the retests the graph names.
godmode session open(if none is already open) thengodmode retest --run- executes the retest commands the archive itself pins to the changed files, one command per runner, and attests the exit code. - Read the raw file only after the graph has been asked. A full-file read is the fallback for a node the graph has no edges for yet, or for understanding logic the graph cannot represent (control flow inside a function) - not the first move on a large file.
Must Not
- Never trust a graph's
impactormust_retestresult; runatlas graph verifyfirst, since a stale snapshot names dependents that no longer hold. - Never run
atlas graph rebuildmerely to erase a failing verify: that destroys the saved snapshot's only record that it had drifted, so fix or explain the drift first instead. - Never read a large file in full before checking the graph: query it first for who calls it and what it would break.
- Never guess at a node id after
unknown_node: true- rebuild or confirm the exact prefixed form instead. - Never skip
atlas closurebefore a change described as low-risk; the point of the graph is that "low-risk" is a claim the archive can check.
Acceptance
atlas graph rebuildderives a fresh graph from the archive and reports its hash, deterministically, from the same archive content every time.atlas closurenames every dependent of a changed file that was not itself changed, before a blast-radius edit is trusted.
If an assertion cannot be proved, report the unmet assertion and the next safe action rather than the outcome it was meant to prove.
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.

