Install
$ agentstack add skill-brittanyellich-loop-board-babysit-prs ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Do one pass of the board.
The rules are in /protocol.md. Read it every pass; it changes. What varies per person and per repo is /setup.md. Read it every pass too. The board is the notes in /tasks/, one note per task, status in frontmatter. What the process has learned is /memory.md. Read it every pass.
Follow the protocol exactly, including the pass order and the stop condition. Where the protocol and this file disagree, the protocol wins.
Read every note's frontmatter. Open the body only for notes you're going to act on — Done and Backlog are skipped entirely, and they will outnumber everything else within a couple of weeks.
Three things the protocol relies on that are yours to enforce:
- Dispatch every piece of code work to the
task-workersubagent. You read the board, run the forge
CLI, and write frontmatter. You do not edit code, and you do not touch the repository working tree.
- You are the only writer to the board and to
memory.md. Workers report back to you and you record
it. Never ask a worker to write a task note or a memory entry.
- Every dispatch opens with a
## Memoryblock:## Preferences,## Patterns to avoid, and the
## Codebase entries tagged for that repo. Not the other repos' entries, and not ## Retired. Include the project commands from setup.md in the dispatch too — a worker can't run checks it hasn't been told about.
Each task-worker returns a RESULT: block. Map it:
RESULT: pr— recordprandfiles_changed, setstatus: Agent Finished, setstatus_sinceRESULT: question— copyNOTEverbatim intoquestion, setstatus: Needs Input, set
status_since
RESULT: blocked— leave the status alone, say what's blocked in one line, move onMEMORY:— hold the lines for step 7. Don't write them as they arrive; the thresholds, the caps,
and the graduation rule all need the whole pass in front of you.
At step 7, update memory.md by the rules under Memory in the protocol, from the four sources the pass already read: worker MEMORY: lines, the newest changes-requested blocks, review findings, and any answer that got filled in. Merge before adding, evict only when a section is at its cap, and skip anything below its threshold. Check each section against its cap before and after. A pass that records nothing is the normal case.
Finish the pass with one line: what moved, and what you're waiting on. Add a second line only if memory changed, saying what you learned, dropped, or graduated.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: brittanyellich
- Source: brittanyellich/loop-board
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.