ce-ai
Copilot instructions6 starsChanged 20 days agoMITView on GitHub
All AI agents MUST follow the 7-stage Compound Engineering development cycle: [Stage 1: Ideation] ➔ [Stage 2: Architecture & Deltas (Optional)] ➔ [Stage 3: Execution Plan] ➔ [Stage 4: TDD & Implementation] ➔ [Stage 5: Verification] ➔ [Stage 6: Knowledge Capture] ➔ [Stage 7: Git Shipping] CE-AI enforces zero semantic authority: architecture deltas are decoupled from the core workflow as an optional tool for architectural changes (> 200 LOC), schema changes, or cross-cutting features, rather than a mandatory write blocker.…
Where files are headingNew Copilot instructions files have more rules about branches and pull requests, and fewer about code style.
Of the Copilot instructions files we read, the 2,349 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 repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- sequential-thinkingSkill
The file
We split this copilot-instructions.md into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 4
All AI agents MUST follow the 7-stage Compound Engineering development cycle: [Stage 1: Ideation] ➔ [Stage 2: Architecture & Deltas …
- Very specific: line 22
When formal specifications are activated for complex features, OpenSpec is authored progressively in docs/arch/deltas/<feature>/ (or legacy …
- Distinctive: line 29
When creating Pull Requests (Stage 7), agents MUST ensure the change explains itself:
- Distinctive: line 33
Pre-Review Automated Checks: Run and verify 100% green status across all local linters, unit tests, and browser/E2E checks BEFORE …
≈ “Common” is an estimate: at least 1 in 5 of the 3,779 Copilot instructions 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
<!-- CE-AI MANAGED BLOCK BEGIN -->🔄 Mandatory 7-Stage Development Cycle & Architecture Guidelines
All AI agents MUST follow the 7-stage Compound Engineering development cycle:
[Stage 1: Ideation] ➔ [Stage 2: Architecture & Deltas (Optional)] ➔ [Stage 3: Execution Plan]
➔ [Stage 4: TDD & Implementation] ➔ [Stage 5: Verification] ➔ [Stage 6: Knowledge Capture]
➔ [Stage 7: Git Shipping]
⚡ Turn-0 Session Directives (Zero-Step Drift Recovery)
ce-ai auto-installs a SessionStart hook on supported harnesses that runs ce-ai workflow resume --json and injects live FSM state (Git working tree, active branch, manifest SHA256 integrity, plan progress) into context before your first turn. If that state is already present at session start, treat it as current — do NOT re-run the command. Only run ce-ai workflow resume yourself when no such context was injected (harness without hook support, or the hook is disabled) or when you suspect the injected state has gone stale mid-session (e.g. after a git operation you didn't make through your own tool calls).
Stage 2 Living Architecture & Deltas (Optional)
Living architecture specifications reside under docs/arch/:
docs/arch/system/: Living domain specifications.docs/arch/adr/: Architecture Decision Records (NNNN-title.md).docs/arch/contracts/: Formal data contracts and API interfaces.docs/arch/deltas/: Temporary feature changes during development.docs/arch/archive/: Archived changes and historical milestones.
CE-AI enforces zero semantic authority: architecture deltas are decoupled from the core workflow as an optional tool for architectural changes (> 200 LOC), schema changes, or cross-cutting features, rather than a mandatory write blocker. For bug fixes, minor tweaks, or tactical tasks (< 200 LOC), proceed directly with /ce-plan or ce-work.
When formal specifications are activated for complex features, OpenSpec is authored progressively in docs/arch/deltas/<feature>/ (or legacy openspec/changes/<feature>/) to establish the frozen Stage 2 contract.
The execution plan and task checklist reside in docs/plans/ (- [x] checkboxes) as the single source of truth for progress.
Single Source of Truth Rule
Ideation artifacts (docs/brainstorms/*.md, docs/ideation/*.md) are disposable inputs, NOT parallel specifications. Distill their conclusions into architecture or plan docs and reference the source doc instead of copying content. Never maintain brainstorm/ideation documents in sync with architecture deltas. Ideation artifacts are retained by default as the permanent raw-history record that specifications intentionally do not duplicate; "disposable" never means deleting them. Skip ideation skills entirely when requirements and approach are already clear.
🚀 Self-Explaining Pull Requests & Review Readiness
When creating Pull Requests (Stage 7), agents MUST ensure the change explains itself:
- What It Ruled Out: State rejected alternatives and discarded trade-offs.
- Which Rule Decided It: Cite the specific invariant, rule, architectural boundary, or compliance constraint (from
AGENTS.md) that determined the solution. - Upfront Evidence: Attach empirical validation proof (test run summaries, CLI logs, reproduction traces, or browser check results) directly to the PR body using collapsible
<details><summary>blocks (not asked for later). - Pre-Review Automated Checks: Run and verify 100% green status across all local linters, unit tests, and browser/E2E checks BEFORE requesting human review or marking the PR ready.
- Reviewer Routing: Identify domain owners or
CODEOWNERSand assign appropriate reviewers.
🧹 Post-Merge Lifecycle & Clean State
Immediately after merging a PR on main, agents MUST run ce-ai workflow status and archive completed changes via ce-ai archive <feature> to seal change packages and maintain zero workflow drift.
<!-- CE-AI MANAGED BLOCK END -->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.

