vanta-context-sync
Skill2 starsChanged 5 months agoMITView on GitHub
Use after completing any Vanta build slice, phase, or structural change (new module, layer, tool, provider, command, env var, API route, or locked decision). Keeps the per-level AGENTS.md context files, README, PRD checkboxes, and project memory accurate and dense so future sessions and subagents don't re-derive context or burn tokens trusting stale docs.
In this repository14 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- vanta-ts/AGENTS.mdAGENTS.md
- .claude/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- docs/CLAUDE.mdCLAUDE.md
- vanta-ts/CLAUDE.mdCLAUDE.md
- customer-discoverySkill
- founder-bottleneckSkill
- gtm-builderSkill
- mvp-scopeSkill
- pmf-diagnosticSkill
- problem-hypothesisSkill
- vanta-context-syncSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 28
Nearest-level AGENTS.md: Update the file map, decisions, and gotchas for the layer you touched — vanta-ts/AGENTS.md for agent work, root …
- Distinctive: line 30
New directory level → new AGENTS.md: If you created a new layer or package, add a AGENTS.md there: one-line role, file map, key decisions, …
- Very specific: line 32
Root AGENTS.md status + maps: Update the ## Status line (current phase, test count, what's next) and the kernel module/API tables if the …
- Distinctive: line 34
README "What works now.": Only if the change is user-facing.
- Distinctive: line 38
Locked decisions: Add to the layer AGENTS.md "Key decisions (don't re-litigate)" section: the choice + a one-line why.
- Very specific: line 40
Project memory: Update MEMORY.md index + project_argo.md with new state, new gotchas, and the next step.
- Distinctive: line 48
If a AGENTS.md exceeds ~120 lines, it's carrying explanation that belongs in code comments or the PRD — trim it.
≈ “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
Vanta Context Sync
Keep Vanta's context files true to the code after every meaningful change. Stale docs cost more tokens than no docs — a subagent that trusts a wrong file map wastes a whole round-trip. This skill is the discipline that stops drift between what the code does and what the docs claim.
When to run
- Finished a build slice or phase (and it verifies — tests pass, typecheck clean)
- Added or removed a module, tool, provider, or directory level
- Changed a command, env var, or API surface
- Locked a design decision
- Hit a gotcha worth recording for next time
Do NOT run for
- Work in progress — wait until the slice actually verifies
- Pure refactors with no structural or decision change
- Trivial edits (typo, comment) that don't change the file map or behavior
Checklist (create a TodoWrite item per step)
- Verify reality first.
lsthe changed directories andgrepthe actual exports, routes, and commands. Update docs from what is, never from memory. A file map listing a deleted file is worse than none. - Nearest-level AGENTS.md. Update the file map, decisions, and gotchas for the layer you touched —
vanta-ts/AGENTS.mdfor agent work, rootAGENTS.mdfor kernel work. One line per file/fact. - New directory level → new AGENTS.md. If you created a new layer or package, add a
AGENTS.mdthere: one-line role, file map, key decisions, how-to-extend, commands. Link up to the parent. - Root AGENTS.md status + maps. Update the
## Statusline (current phase, test count, what's next) and the kernel module/API tables if the kernel changed. - README "What works now." Only if the change is user-facing. Keep the run commands runnable — actually run them.
- PRD checkboxes (
docs/prd.md). Tick the phase done-criteria you met. If scope shifted, note it explicitly — don't rewrite the roadmap silently. - Locked decisions. Add to the layer AGENTS.md "Key decisions (don't re-litigate)" section: the choice + a one-line why. This is what stops future re-debate.
- Project memory. Update
MEMORY.mdindex +project_argo.mdwith new state, new gotchas, and the next step. Convert relative dates to absolute. - Self-check. Re-read each edited file: does the map match
ls? Do the commands run? Any reference to a renamed or deleted thing? Fix inline.
Density rules (non-negotiable — token economy is the whole point)
- Tables and one-liners over prose. No "this section describes…" filler.
- Decisions get a why; facts don't.
- If a AGENTS.md exceeds ~120 lines, it's carrying explanation that belongs in code comments or the PRD — trim it.
- Global
~/.Codex/AGENTS.mdconventions are inherited, never repeated. - The test: could a fresh subagent do the next slice from these files without opening more than 2-3 source files? If not, the docs are incomplete. If they're reading docs longer than the code, they're bloated.
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.

