kontext-tasks
Skill1 starsChanged 3 months agoMITView on GitHub
Run work through Kontext's task loop. Trigger when the user starts a non-trivial task, asks to track or resume work, reports progress or a blocker, finishes something, or asks what to work on next. Do not trigger for one-off questions with no durable outcome.
In this repository9 other files it gives its agents.
- catch-upSkill
- kontext-contextSkill
- kontext-memorySkill
- kontext-sharingSkill
- log-decisionSkill
- recallSkill
- saveSkill
- save-skillSkill
- statusSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Call get_my_context and search_context for anything relevant to the task.
- Very specific: line 16
If a matching open task exists, adopt it.
- Very specific: line 18
If the task belongs to a project, link it with parentId.
- Distinctive: line 23
Update the task at real milestones only: a decision, a blocker, a finished phase.
- Very specific: line 25
When a decision is made or durable context emerges, save it (save_to_kontext or manage_document) and keep the task record current.
- Distinctive: line 27
If work stalls on something external, record the blocker so the next session can pick it up.
- Very specific: line 39
When the user asks what to work on next: list open tasks with list_tasks, respect blockedByIds and priorities, and present a short, …
≈ “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.
Files
Kontext task loop
Carry work through Kontext end to end: check what already exists, track the
task, keep it current, close it out, then decide what is next. Requires the
Kontext MCP tools (see kontext-context).
1. Before work
- Call
get_my_contextandsearch_contextfor anything relevant to the task. Read what is relevant before planning. - If a matching open task exists, adopt it. Otherwise create one with
manage_taskand set itin_progress. - If the task belongs to a project, link it with
parentId. If it depends on or blocks other tasks, record that withmanage_task_relationship.
2. During work
- Update the task at real milestones only: a decision, a blocker, a finished phase. Do not spam status notes.
- When a decision is made or durable context emerges, save it
(
save_to_kontextormanage_document) and keep the task record current. - If work stalls on something external, record the blocker so the next session can pick it up.
3. After work
- Mark the task
donewith an outcome summary in its description. - Save the durable learnings that will matter in another conversation.
- Propose the two or three best next steps from the user's open tasks, then stop and wait for the user to choose. Never start the next task unprompted.
What next
When the user asks what to work on next: list open tasks with
list_tasks, respect blockedByIds and priorities, and present a short,
concrete recommendation with reasons — do not begin it without confirmation.
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.

