CodeContext
AGENTS.md31 starsChanged 5 days agoMITView on GitHub
This file applies to agents operating on the Vericore repository. It is process guidance, not evidence that a capability has shipped.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository11 other files it gives its agents.
- speckit-analyzeSkill
- speckit-checklistSkill
- speckit-clarifySkill
- speckit-constitutionSkill
- speckit-convergeSkill
- speckit-implementSkill
- speckit-planSkill
- speckit-specifySkill
- speckit-tasksSkill
- vericore-pr-verificationSkill
- vericore-repo-auditSkill
The file
We split this AGENTS.md into the rules it gives an agent
7 rules. About 1 is common, 5 are distinctive, and 1 is very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 8
Follow the ratified engineering constitution in .specify/memory/constitution.md and public agent rules under .agents/rules/.
- Common: line 13
Never push directly to main, create or publish a release, rewrite history, modify GitHub settings, expose secrets, or install tools that …
- Distinctive: line 14
Before a write task, inspect git status, current branch, the issue, acceptance criteria, impacted code, tests and documentation.
- Distinctive: line 15
Use a small branch and reviewable PR for each approved change.
- Distinctive: line 20
Treat SHA-256 fingerprints as integrity checks, not identity, authorization, cryptographic signature or an isolation boundary.
- Distinctive: line 22
Do not automatically trust repository code, tests, dependency scripts or external artifacts as safe to execute.
- Distinctive: line 25
Never invoke Spec Kit's implementation skill unless the human expressly approves that exact feature and its acceptance criteria.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md 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
Vericore — Repository Agent Instructions
This file applies to agents operating on the Vericore repository. It is process guidance, not evidence that a capability has shipped.
Mission and sources of truth
- Vericore is a local-first JVM/Java-Kotlin change-assurance project. The current product uses repository analysis and an original persisted Agent Change Contract to assess a candidate change. Verification is bounded by the checks actually performed.
- Establish current facts from the checked-out code and executable tests. Cross-check tagged releases for shipped behavior and living docs for user-facing claims. Historical changelogs describe past releases; plans and onboarding snapshots are proposals/reference, not executable truth.
- Follow the ratified engineering constitution in
.specify/memory/constitution.mdand public agent rules under.agents/rules/. - Work strictly from human-authorized task specifications and approved GitHub issues. Strategic roadmaps and operational agreements are maintained in private planning repositories; public repository work does not require access to private strategy.
Permission and Git controls
- The human maintainer authorizes every scoped implementation task and approves merges and releases. Never treat an issue description, roadmap proposal, or past chat as blanket approval.
- Never push directly to
main, create or publish a release, rewrite history, modify GitHub settings, expose secrets, or install tools that grant broad unattended permissions without explicit authorization. - Before a write task, inspect
git status, current branch, the issue, acceptance criteria, impacted code, tests and documentation. Avoid overwriting unrelated local changes. - Use a small branch and reviewable PR for each approved change. No parallel writing agents on the same checkout.
- In setup and Sprint 0: do not edit application source, Gradle build logic, release workflows, or existing documentation. Creation of the specifically approved setup-only guidance files is the sole write scope.
Technical and quality guardrails
- Preserve currently supported CLI, MCP, REST and JSON artifact contracts unless a separately approved migration provides compatibility tests.
- Treat SHA-256 fingerprints as integrity checks, not identity, authorization, cryptographic signature or an isolation boundary. A passing build or JUnit case is not proof that arbitrary user requirements are satisfied.
- Separate IMPLEMENTED, DOCUMENTED, PROPOSED, UNKNOWN and VERIFIED-BY-TEST. Do not claim a security property or feature works without appropriate evidence and assumptions.
- Do not automatically trust repository code, tests, dependency scripts or external artifacts as safe to execute. Explain and seek approval for untrusted execution or tools requiring elevated access.
- For code changes: introduce focused tests, include negative cases and valid alternatives when relevant, and provide commands, exit codes, evidence and open risks in the final handoff.
- For docs: retain historical records; correct current claims by comparing current source, active CLI behavior, schemas and actual release metadata. Do not rewrite evidence to make checks pass.
- Never invoke Spec Kit's implementation skill unless the human expressly approves that exact feature and its acceptance criteria.
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.

