skills.rest
EveryInc
EveryIncOFFICIAL
Explore all skills by this author →
WebsiteGitHub

ce-sweep

Sweep Slack and GitHub feedback sources, acknowledge items, verify fixes, and emit a rolling plan.

24.7k|2.0k|Updated Oct 9, 2025
Search Tags:#Slack#Automation#fix verification#feedback triage#gitHub issues#State management#recording analysis
One-click install
npx skills add https://github.com/EveryInc/compound-engineering-plugin --skill ce-sweep
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ce-sweep
Source: https://github.com/EveryInc/compound-engineering-plugin/tree/main/skills/ce-sweep
Command: npx skills add https://github.com/EveryInc/compound-engineering-plugin --skill ce-sweep

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Product feedback scattered across Slack channels and GitHub Issues goes unacknowledged, recordings go unanalyzed, and claimed fixes go unverified. This Skill runs a repeatable sweep that acknowledges every new item at its source, analyzes attached recordings, verifies that claimed fixes actually merged to the default branch, and folds open items into a single rolling plan.

Core Features & Use Cases

  • Multi-source feedback ingestion: Fetch new items from Slack channels and GitHub Issues (email experimental) using per-source cursors, with a deterministic state engine as the only writer of sweep state.
  • Source-side acknowledgment and close-out: Apply configured emoji reactions or labels to acknowledge items and close them out once a fix is verified merged, gated by per-source standing approval and an acknowledgment circuit breaker.
  • Recording analysis and fix verification: Download attached media to scratch, analyze recordings via sub-agents, and validate claimed fix refs (PR numbers or commit SHAs) against the default branch before closing items.
  • Rolling plan output: Reconcile a stable feedback-sweep-plan.md with stable requirement IDs, drained closed items, and outstanding questions, ready to hand off to an lfg run.
  • Use Case: A team schedules a nightly non-interactive sweep that acks new Slack feedback, analyzes screen recordings, confirms yesterday's PR fixed a reported bug, and commits an updated plan for the morning triage.

Quick Start

Run the ce-sweep skill to set up your feedback sources interactively on first run, then sweep them for new items and produce an updated feedback plan.

Frequently Asked Questions about ce-sweep

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up automated feedback triage from Slack and GitHub Issues?▼

Run ce-sweep interactively once to complete the first-run interview, which captures each source's type, target, acknowledgment action, approval, and sensitivity into config.local.yaml. Subsequent runs fetch new items, acknowledge them at the source, and update the rolling plan.

How does the sweep verify that a reported bug fix actually shipped?▼

For each fix_pending item, the sweep validates the claimed fix ref as a bare PR number or commit SHA, then checks merge status with gh pr view or git merge-base against the default branch. Only verified items are closed with fix_ref, verified_merge_sha, and verified_at evidence.

Can the feedback sweep run on a schedule without prompting?▼

Yes, pass mode:non-interactive when registering the invocation with a scheduler. Non-interactive runs never prompt; ambiguous decisions and circuit-breaker trips are deferred into the plan's Outstanding Questions section instead.

What happens if two sweep runs overlap on the same repository?▼

A single-writer lease in the state file serializes runs: a second run gets LOCKED, records an aborted-locked outcome, and exits. Expired leases are reclaimed after the TTL, and shared-branch mode makes the lease a repo-wide mutex via push gating.

How is sensitive feedback content handled in state and plans?▼

Sources marked sensitive have item body and quote fields dropped by the state engine at write time, so redacted content never reaches disk. Plans reference such items by id, title, and url only, with the quote replaced by a withheld-content marker.

What are the limitations of the email feedback source?▼

Email sources are experimental and read-only: they have no source-side acknowledgment or close-out actions, so items are tracked only in state and recorded with approved set to false. Slack and GitHub Issues are the fully supported source types.

Related Skills

Explore More Product & Management Skills→
redhat-etredhat-et
9

context-fetcher

Search GitHub, Confluence, and Slack for incident-related configuration, documentation, and messages.

Official
Intermediate
ContentsUSContentsUS

gogogo

Execute implementation plans from GitHub issues into pull requests.

Official
Advanced
jikig-aijikig-ai
12

schedule

Generate pinned GitHub Actions cron workflows that run Soleur skills.

Official
Advanced
antondelafuenteantondelafuente

triage-feedback

Automate triage and disposition of product issues with defined labels.

Community
Advanced