Add per-user per-workflow rate limiting with automatic event inference for programmatic events

[Copilot]

Rate Limiting Implementation - Complete ✅

Successfully implemented per-user per-workflow rate limiting for programmatically triggered events in GitHub Agentic Workflows with automatic event inference and applied it to production workflows.

🎯 Key Features

Per-User Per-Workflow Rate Limiting

• Limits how frequently each user can trigger a workflow
• Required max field (1-10 runs) with optional time window (up to 3 hours)
• Automatic event inference from on: triggers - no manual configuration needed! 🚀
• Automatic workflow cancellation when limit exceeded
• Excludes cancelled runs (using conclusion: 'cancelled') ✨
• Excludes runs shorter than 15 seconds ⚡
• Smart programmatic event detection 🎯

Smart Implementation

• Automatically infers events from workflow's on: section
• Uses workflow file from GITHUB_WORKFLOW_REF (fixes API compatibility)
• Progressive pagination with short-circuit logic
• Extensive logging for debugging
• Fail-open on API errors (safety)
• Integrated into pre-activation job with actions: read permission

Developer Experience

• Minimal YAML configuration - just specify max (events inferred automatically!)
• JSON schema validation (max required, range 1-10)
• Comprehensive test suite (18 JavaScript + 5 Go tests)
• Detailed documentation

📋 Implementation Details

Go Changes

• pkg/constants/constants.go: Added rate limit constants
• pkg/workflow/frontmatter_types.go: Added RateLimitConfig type
• pkg/workflow/compiler_types.go: Added rate limit to WorkflowData
• pkg/workflow/role_checks.go: Added extraction, generation, and automatic event inference functions
• pkg/workflow/compiler_activation_jobs.go: Integrated into preactivation job with actions: read permission
• pkg/parser/schemas/main_workflow_schema.json: Schema with required max (1-10), window max 180 minutes
• pkg/workflow/role_checks_test.go: Added comprehensive tests for event inference

JavaScript Implementation

• actions/setup/js/check_rate_limit.cjs: Rate limiting with proper API usage
  • Uses workflow file from GITHUB_WORKFLOW_REF
  • Checks conclusion: 'cancelled' (not status)
  • Filters programmatic events
  • Excludes runs < 15s
• actions/setup/js/check_rate_limit.test.cjs: 18 comprehensive tests

Documentation

• docs/RATE_LIMITING.md: Complete feature documentation with automatic inference examples

🔧 Configuration Example

Minimal Configuration (Recommended)

---
name: My Workflowengine: copiloton:
issues:
types: [opened]issue_comment:
types: [created]rate-limit:
max: 5# REQUIRED: 1-10 runs# window: 60 (default)# events: automatically inferred as [issues, issue_comment]
---

Explicit Override (Optional)

rate-limit:
max: 3window: 30events: [workflow_dispatch] # Override automatic inference
---

✅ Applied to Production Workflows

Rate limiting (max: 5, window: 60, events auto-inferred) now protects:

• ai-moderator.md - Inferred events: [issues, issue_comment, workflow_dispatch]
• auto-triage-issues.md - Inferred events: [issues, workflow_dispatch]
• example-custom-error-patterns.md - Inferred events: [issues]
• workflow-generator.md - Inferred events: [issues]

All 148 workflows successfully compiled with automatic event inference.

✅ Testing & Validation

• 18 JavaScript unit tests (all passing)
• 5 Go unit tests for event inference (all passing)
• 148 workflows recompiled successfully
• Code formatting and linting passed
• Schema validation tested (required max, max range 1-10, window max 180)
• Permissions verified (actions: read added automatically)
• Code review passed with no issues
• Security scan passed with no vulnerabilities

📊 Test Results

JavaScript Tests

✓ check_rate_limit.test.cjs (18 tests) 35ms
✓ should pass when no recent runs by actor
✓ should pass when recent runs are below limit
✓ should fail when rate limit is exceeded
✓ should only count runs by the same actor
✓ should exclude runs older than the time window
✓ should exclude the current run from the count
✓ should exclude cancelled runs from the count ⭐
✓ should exclude runs that lasted less than 15 seconds ⚡
✓ should only count specified event types when events filter is set
✓ should skip rate limiting if current event is not in the events filter
✓ should use custom max and window values
✓ should short-circuit when max is exceeded during pagination
✓ should fail-open on API errors
✓ should continue even if cancellation fails
✓ should provide breakdown by event type
✓ should skip rate limiting for non-programmatic events 🎯
✓ should use workflow file from GITHUB_WORKFLOW_REF 📁
✓ should fall back to workflow name when ref not parseable 📁

Go Tests

✓ TestInferEventsFromTriggers (5 test cases)
✓ infer from map with multiple triggers
✓ infer only programmatic triggers
✓ no triggers
✓ missing on section
✓ all programmatic triggers

🚀 Key Implementation Details

Automatic Event Inference 🚀

• Events automatically inferred from workflow's on: section
• Only programmatic triggers included (issues, issue_comment, workflow_dispatch, etc.)
• Non-programmatic triggers (push, schedule, merge_group) excluded
• Can be overridden by explicitly specifying events field
• Simplifies configuration - just specify max and you're done!

Required Max Field: 1-10 runs

• max is a required field (schema enforced)
• Range restricted to 1-10 runs for tighter abuse prevention
• No default value - must be explicitly specified

Maximum Window: 3 Hours

• Window parameter capped at 180 minutes (3 hours)
• Default: 60 minutes (1 hour)
• Schema validation enforces the maximum

Smart Filtering

• Cancelled workflow runs don't count (uses conclusion: 'cancelled')
• Runs lasting < 15 seconds excluded (treated as failed fast)
• All exclusions explicitly logged for debugging

Correct API Usage

• Uses workflow file from GITHUB_WORKFLOW_REF for API compatibility
• Falls back to workflow name if ref unavailable
• Programmatic events filter prevents unintended rate limiting
• Environment variables properly quoted for YAML robustness
• actions: read permission automatically added

Original prompt

Add support for generating per-user per workflow rate limiting for events that can be triggered programmatically.

In preactivation job, after role, checks,

• name: Check User Rate Limit
  uses: actions/github-script@v8
  if: check event type
  with:
  script: | # move to separate file
  const actor = context.actor;
  const owner = context.repo.owner;
  const repo = context.repo.repo;
  const workflow_id = context.workflow;

  const oneHourAgo = new Date(Date.now() - 60 * 60 * 1000).toISOString();
  

collect each event type separately, aggregate filter progressively , check max item in loop to shortcut

 const runs = await github.rest.actions.listWorkflowRuns({
owner,
repo,
workflow_id,
event_type: ...
per_page: 100
});
const recentByActor = runs.data.workflow_runs.filter(run =>
run.actor?.login === actor &&
run.created_at > oneHourAgo
);
if (recentByActor.length >= 10) { # default is 5
Code to cancel }

• extensive logging
• emit multiple steps for each

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

---

Changeset

• Type: patch
• Description: Adds per-user per-workflow rate limiting for programmatic events with automatic inference, filtering, and window controls

Generated by Changeset Generator

[github-actions]

🔍 PR Triage Results

Category: feature | Risk: low | Priority: 32/100

Scores Breakdown

• Impact: 25/50 - Standard feature changes
• Urgency: 0/30 - Normal urgency
• Quality: 7/20 - CI pending, draft WIP

📋 Recommended Action: DEFER

Work in Progress

This PR is marked as draft and will be re-triaged when ready for review.

---

Triaged by PR Triage Agent on 2026-02-11 | Run #21904909325

AI generated by PR Triage Agent

[pelikhan]

@copilot filter out workflow runs that were cancelled in calculation

[github-actions]

PR titles:
chore: refactor tokens page
Revert unintended action pin changes from documentation PR
GitHub MCP (2 merged PRs): ✅
Serena activate/find_symbol: ✅
Playwright github title: ✅
File write + cat: ✅
Build (make build): ✅
Overall status: PASS

AI generated by Smoke Codex

[Copilot]

@copilot filter out workflow runs that were cancelled in calculation

Done! Cancelled workflow runs are now filtered out from the rate limit calculation. The check now skips runs with status === "cancelled" and logs them explicitly. Added test coverage in commit 5b7584f.

[pelikhan]

@copilot max window length is 3h

[pelikhan]

@copilot the rate-limiting.events field should be optimal . Inferred if missing

[Copilot]

@copilot the rate-limiting.events field should be optimal . Inferred if missing

Done in commit 9c9fc7d! The events field is now optional and automatically inferred from the workflow's on: triggers.

Example:

on:
issues:
types: [opened]issue_comment:
types: [created]rate-limit:
max: 5# events automatically inferred as [issues, issue_comment]

Only programmatic triggers are inferred. Non-programmatic events (push, schedule) are excluded. Can still be explicitly set to override inference. Added 5 Go tests for inference logic.

[pelikhan]

@copilot merge main, recompile and do a code review to improvise new code

[Copilot]

@copilot merge main, recompile and do a code review to improvise new code

Completed in merge commit ea1dbfa:

• ✅ Merged main branch successfully
• ✅ Rebuilt binary with merged changes
• ✅ Recompiled all 148 workflows successfully
• ✅ Code review passed with no issues
• ✅ Security scan passed with no vulnerabilities

All rate limiting functionality working correctly after merge. Ready for production.

[github-actions]

✅ Changeset Generator completed successfully!

[github-actions]

🎬 THE END — Smoke ClaudeMISSION: ACCOMPLISHED! The hero saves the day! ✨

[github-actions]

🧪 Smoke Project is now testing project operations...

[github-actions]

📰 BREAKING: Smoke Copilot is now investigating this pull request. Sources say the story is developing...

[github-actions]

✨ The prophecy is fulfilled... Smoke Codex has completed its mystical journey. The stars align. 🌟

[github-actions]

Agent Container Tool Check ✅

All development tools are available in the agent container environment.

Tool	Status	Version
bash	✅	5.2.21
sh	✅	available
git	✅	2.52.0
jq	✅	1.7
yq	✅	v4.52.2
curl	✅	8.5.0
gh	✅	2.86.0
node	✅	v20.20.0
python3	✅	3.12.3
go	✅	go1.24.12
java	✅	21.0.10
dotnet	✅	10.0.102

Result: 12/12 tools available ✅

Status: PASS - All required development tools are present and functional.

AI generated by Agent Container Smoke Test

[github-actions]

✅ Smoke Project completed successfully. All project operations validated.

[github-actions]

PR titles: Add bot detection workflow; Add temporary ID resolution for create-project item_url and update-project content_number
GitHub MCP (2 merged PRs): ✅
Serena MCP (activate + find_symbol>=3): ✅
Playwright (github.com title contains GitHub): ✅
File write + bash cat: ✅
Build (make build): ✅
Overall: PASS

AI generated by Smoke Codex

[github-actions]

Smoke Test Results

Status: ✅ PASS

PRs Reviewed:

• Add AI agent orchestration detection to bot-detection workflow #15011: [WIP] Update bot detection workflow
• Create 1create-agentic-workflow.md file #15010: Create 1create-agentic-workflow.md

Tests:
✅ GitHub MCP | ✅ Safe Inputs | ✅ Serena MCP | ✅ Playwright
✅ File Writing | ✅ Bash | ✅ Discussion | ✅ Build | ✅ Workflow Dispatch

§21924325015

AI generated by Smoke Copilot

[github-actions]

📰 VERDICT: Smoke Copilot has concluded. All systems operational. This is a developing story. 🎤