attention-frontier
Skill69 starsChanged 11 days agoMITView on GitHub
Maintain a person's prioritized follow-up frontier across parallel projects and agent conversations. Use to consolidate pending decisions, reviews, and coordination into a current workbench with brief change notifications.
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
Scope, incremental evidence, reconciliation: Start from the user's working scope and existing task map.
- Distinctive: line 54
Read presentation.md when creating or updating it.
- Distinctive: line 57
Use the independent view across matters to recommend the most useful next move and briefly explain why.
- Distinctive: line 70
Durable context, honest freshness: Keep a compact continuation record: scope and source coverage, matter-to-source mappings, stable read …
- Distinctive: line 77
Run one pass unless ongoing checks were requested.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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.
Files
Attention Frontier
Reduce the person's cost of surveying work, chasing progress, and recovering context while keeping important unfinished work accounted for.
Ground
Scope, incremental evidence, reconciliation. Start from the user's working scope and existing task map. Establish an initial baseline, then discover new work and read source increments. Recent activity is a discovery window, not an expiry date for older unfinished work.
Group related conversations by business matter; retain distinct user actions and unfinished side or parallel tasks, including those within one conversation. Recover the main objective and relevant backstory from source context when a recent turn only describes a local step. Reconcile candidates against current user instructions, related conversations, and relevant delivery evidence. Later resolutions must retire stale requests. Distinguish an agent's finished turn, compilation, tests, delivered artifacts, and business acceptance. Preserve the evidence level of each claim.
Select and order
Human action, ownership, business priority. Identify the next decision, review, hands-on action, or coordination that actually needs the person. Ordinary authorized execution stays with its owner. When the person's action is done and ownership continues elsewhere, move the matter to observation; bring it back when a new user action becomes ready. Reading a report alone does not complete an action.
Proactively investigate missing evidence. Remaining uncertainty belongs in a clearly labeled area with its gap and possible impact; elevate it to a user action only when the person is needed to resolve it.
Rank by explicit user priorities, deadlines and business impact, delay cost, dependencies released, and waiting time. Preserve relative order where these have not materially changed. Use judgment rather than invented deadlines, fixed scores, quotas, or Top-N limits. Keep all unfinished matters accessible, including work progressing independently and external waits.
Publish
Full coverage, selective emphasis, easy re-entry. Maintain an HTML workbench at a stable entry point by default; adapt the medium to the environment. Read presentation.md when creating or updating it. In Codex App, also read codex.md.
Use the independent view across matters to recommend the most useful next move and briefly explain why. Lead with that recommendation when user action is ready. Orient the reader to each matter's purpose, background, and related work before showing the next step, next owner, reason, source entry, and a short suggested prompt the person can send to its owning conversation.
Reflect priority, overdue status, and uncertainty whenever they change. Keep polling replies to a sentence or two about meaningful changes plus the entry link; stay quiet when nothing substantive changed. An unresolved item remains visible without requiring repeated identical messages.
Continue
Durable context, honest freshness. Keep a compact continuation record: scope and source coverage, matter-to-source mappings, stable read positions, unresolved and resolved user actions, owners, check conditions, and the previous reported state. Choose the storage format for the environment. Preserve resolution provenance so old requests do not reappear. On a failed read, retain the last known state and its original verification time; mark the gap.
Run one pass unless ongoing checks were requested. For ongoing work, use the host's native scheduler with the agreed cadence and time zone, reuse an existing matching schedule, and report unavailable capabilities. End the turn between scheduled runs. Stop this watcher's continuation when it is canceled.
This role observes, prioritizes, presents, and notifies. Leave task execution with existing owners; messaging them or restarting their work requires separate authorization.
In this repository28 other files it gives its agents.
- antigravity-coding-agentSkill
- audio-transcriptionSkill
- brainstormSkill
- build-verifierSkill
- claude-code-coding-agentSkill
- codex-async-followupSkill
- codex-coding-agentSkill
- codex-session-controllerSkill
- codex-session-namingSkill
- cursor-coding-agentSkill
- debuggingSkill
- deslopSkill
- explainSkill
- explore-and-planSkill
- faster-learning-coachSkill
- frontend-harness-slidesSkill
- generate-commit-messageSkill
- git-task-workspaceSkill
- grok-coding-agentSkill
- home-config-syncSkill
- llm-wiki-captureSkill
- local-file-organizerSkill
- long-task-controlSkill
- parallel-goal-workflowsSkill
- reverse-engineerSkill
- review-and-improveSkill
- tddSkill
- x-twitter-readerSkill
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.

