agentleFS
Fei2-LabsFei2-Labs/skill-genie/skills/dev-inbox/SKILL.md

dev-inbox

Skill12 starsChanged 49 days agoMITView on GitHub

Triage and persist work outside the explicit active objective. Use at message intake when a message contains multiple requests, an item is unrelated or deferred, the user asks to remember, track, log, or handle something later, or invokes dev-inbox. After resume or context compaction, audit requested-but-unrecorded items before continuing.

1 other project uses a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

23 rules. About 0 are common, 19 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈19

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run this gate before acting on every multi-request message:

    line 20
  2. Distinctive:

    After a resume, handoff, or context compaction, audit the recovered context for any explicit recording request that lacks a confirmed …

    line 29
  3. Very specific:

    Treat an explicit request to record or defer an item as consent.

    line 76
  4. Distinctive:

    Check for an existing related record.

    line 79
  5. Distinctive:

    Ask one focused question only when deduplication is genuinely ambiguous.

    line 83
  6. Distinctive:

    When you identify something that doesn't belong to the current task, respond with a concrete suggestion — not an open question:

    line 87
  7. Distinctive:

    If a resume summary says the user requested an item be recorded but no destination or confirmation exists, treat it as unresolved.

    line 117
  8. Distinctive:

    Before creating a new record, check if a related one already exists:

    line 205
  9. Very specific:

    Memory: Check for [TODO] entries with similar content

    line 209
  10. Distinctive:

    If a match is found → append as a sub-item or checklist entry.

    line 211
  11. Distinctive:

    If uncertain → ask: "This looks related to [existing item].

    line 212
  12. Distinctive:

    When using GitHub Issues, apply these labels (create if they don't exist):

    line 218
  13. Very specific:

    If gh is not installed but a GitHub remote exists, output:

    line 233
  14. Distinctive:

    Do not interrupt the user's flow with long explanations.

    line 253
  15. Distinctive:

    Do not ask open-ended questions like "What do you want to do with this?" — propose a concrete action.

    line 254
  16. Distinctive:

    Do not ask for confirmation after an explicit recording or deferral request.

    line 255
  17. Distinctive:

    Do not create records that can't be found later.

    line 256
  18. Distinctive:

    Do not duplicate: always check for existing related records first.

    line 257
  19. Distinctive:

    Do not use this skill for things that ARE part of the current task — just do them.

    line 258
  20. Distinctive:

    Do not treat every topic in the conversation, open files, or git diff as part of the active objective.

    line 259
  21. Distinctive:

    Do not resume active work while a recovered explicit recording request remains unresolved.

    line 260
  22. Distinctive:

    Do not generate a file and leave it disconnected.

    line 261
  23. Very specific:

    handoff-receiver: When resuming, check for pending inbox items as part of context loading.

    line 268

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Dev Inbox

Triage anything that comes up during a session — bugs, features, improvements, fleeting ideas — and route it to the right place so it is never lost and always discoverable by future sessions.

This skill works in any context: software development, writing, design, or any task.


Intake Gate

Run this gate before acting on every multi-request message:

  1. Identify the explicit active objective from the active task artifact or the user's latest scoped instruction.
  2. Split every bullet, sentence-level request, and aside into atomic items.
  3. Classify each item as active-objective work, a blocking prerequisite, or unrelated/deferred inbox work.
  4. Route explicit inbox items before returning to active-objective work.

Conversation history, open files, and git diff are context, not proof that an item belongs to the active objective. Mere presence in the current conversation does not make a topic current-task work.

After a resume, handoff, or context compaction, audit the recovered context for any explicit recording request that lacks a confirmed destination. Persist it before continuing the active objective.


Decision Tree

User says something
│
├─ Is it part of the current task?
│   (Current task = active task artifact or latest explicitly scoped objective)
│   ├─ YES → Do it directly. Stop here.
│   └─ NO ↓
│
├─ Does it block the current task?
│   ├─ YES → Handle it first, then return to current task.
│   └─ NO ↓
│
├─ Priority?
│   ├─ HIGH — would cause data loss, money loss, security issue, or blocks others
│   ├─ NORMAL — clearly needs doing, but not urgent
│   └─ LOW — nice-to-have, fleeting thought, cosmetic
│
└─ Record it + ensure future discoverability

Classification

Assign one type and one priority:

TypeMeaningExample
fixSomething existing is broken or wrongBug, incorrect content, wrong behavior
addSomething new is neededFeature, new section, new capability
improveWorks but could be betterBetter wording, cleaner UI, performance
ideaFleeting thought, maybe later"What if we also..."

Priority: high / normal / low


Interaction Pattern

Explicit request

Treat an explicit request to record or defer an item as consent. This includes phrases such as "记一下", "以后再说", "log this", "track this", and direct dev-inbox invocation.

  1. Classify the item.
  2. Check for an existing related record.
  3. Persist or merge it without asking for confirmation again.
  4. Report the record location, then return to the active objective.

Ask one focused question only when deduplication is genuinely ambiguous.

Proactive inference

When you identify something that doesn't belong to the current task, respond with a concrete suggestion — not an open question:

This doesn't seem related to the current task. I suggest recording it as:

Type: fix | Priority: normal Title: Receipt is not generated after order submission

Confirm?

The user responds with one word (yes/no/adjust). Then execute.


Regression Examples

Multi-topic explicit request

User:

继续修 Browser UI。另外打开会话特别慢,用 Dev-inbox 记一下,后面再说。

Expected order:

  1. Split the Browser UI work and session-opening performance problem into separate items.
  2. Deduplicate and persist the performance item first because the user explicitly requested recording.
  3. Report where it was recorded.
  4. Return to the Browser UI objective.

Resume recovery

If a resume summary says the user requested an item be recorded but no destination or confirmation exists, treat it as unresolved. Deduplicate and persist it before resuming the active objective.


Where to Record

Detect the environment and pick the best destination. The goal is future discoverability — the record must surface in a future session without the user remembering it exists.

Priority order:

  1. GitHub remote + gh CLI available
    • Check: git remote get-url origin 2>/dev/null && command -v gh
    • Action: gh issue create with title, label, and body
    • Why discoverable: Agent can gh issue list --state open --label <type>
    • Merge logic: Search open issues with same label + similar title keywords. If found → append checklist item. If unsure → ask user.
  2. Agent memory system available (Devin memories, Claude memory, etc.)
    • Action: Write to memory with [TODO] prefix and type/priority metadata
    • Why discoverable: Automatically loaded on next session start
  3. Project directory exists
    • Action: Append to TODO.md in project root (create if absent)
    • Format: Grouped by type, each item has priority tag
    • Why discoverable: Agent should read TODO.md at session start
  4. None of the above
    • Action: Output formatted content for the user to place manually
    • Format: Ready to paste anywhere

After recording, confirm:

Recorded: [type] [title]
Location: [where it was saved]
Discovery: [how a future session will find it]

Record Templates

Adapt detail level to priority — lower priority = lighter format.

fix (high/normal)

## Problem
[What is broken / wrong]

## Expected
[What should happen]

## Context
[Where/when discovered, related task if any]

fix (low)

- [One-line description of what's wrong]

add

## What
[What to add]

## Why
[Why it matters]

improve

- [What it is now] → [What it should be]

idea

- [One sentence]

Merge Logic

Before creating a new record, check if a related one already exists:

  • GitHub Issues: gh issue list --state open --label <type> — scan titles for keyword overlap
  • TODO.md: Scan same type section for similar items
  • Memory: Check for [TODO] entries with similar content

If a match is found → append as a sub-item or checklist entry. If uncertain → ask: "This looks related to [existing item]. Add to it, or create separate?"


GitHub Issue Labels

When using GitHub Issues, apply these labels (create if they don't exist):

gh label create "fix" --color "d73a4a" --description "Something is broken" 2>/dev/null
gh label create "add" --color "0075ca" --description "New feature or content" 2>/dev/null
gh label create "improve" --color "a2eeef" --description "Enhancement to existing" 2>/dev/null
gh label create "idea" --color "e4e669" --description "Exploration, maybe later" 2>/dev/null
gh label create "high" --color "b60205" --description "High priority" 2>/dev/null
gh label create "low" --color "c5def5" --description "Low priority" 2>/dev/null

No-gh Fallback

If gh is not installed but a GitHub remote exists, output:

I can't create the issue automatically (gh CLI not found).
Here's the issue ready to paste:

Title: [title]
Labels: [type], [priority]
Body:
---
[formatted body]
---

Create it at: [repo URL]/issues/new

Anti-patterns

  • Do not interrupt the user's flow with long explanations. Keep triage to 2–3 lines.
  • Do not ask open-ended questions like "What do you want to do with this?" — propose a concrete action.
  • Do not ask for confirmation after an explicit recording or deferral request; the request is consent.
  • Do not create records that can't be found later. Always state the discovery path.
  • Do not duplicate: always check for existing related records first.
  • Do not use this skill for things that ARE part of the current task — just do them.
  • Do not treat every topic in the conversation, open files, or git diff as part of the active objective.
  • Do not resume active work while a recovered explicit recording request remains unresolved.
  • Do not generate a file and leave it disconnected. The record must be in a system the agent will check.

Relationship to Other Skills

  • session-handoff / close-loop: At session end, mention count of items recorded this session (e.g., "Recorded 3 items to dev-inbox this session"). No need to re-save — items are already persisted.
  • handoff-receiver: When resuming, check for pending inbox items as part of context loading.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.