initiative
Skill59 starsChanged 12 days agoMITView on GitHub
Create or continue multi-task development and testing initiatives with a shared Serena plan, acceptance checks, and one task per conversation. Use for work that needs task handoffs across conversations, not routine single-task changes.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- qyndex/claude-inittheir own version, first seen Oct 13, 2026
- 4erkashin/skillstheir own version, first seen Sep 7, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- 4erkashin/skills★ 2same name
- qyndex/claude-init★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- clawkerd/AGENTS.mdAGENTS.md
- controlplane/AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- test/AGENTS.mdAGENTS.md
- .github/copilot-instructions.mdCopilot instructions
- agent-filesSkill
- audit-memorySkill
- cli-outputSkill
- dev-checksSkill
- firewall-uatSkill
- monitoring-checksSkill
- store-editorSkill
- test-hunterSkill
- wizard-commandsSkill
- writing-testsSkill
- code-reviewSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 10 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈10
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
Read the project instructions, relevant code, and existing plans.
- Distinctive: line 22
Use assets/initiative.md to create the memory.
- Very specific: line 26
Follow .serena/memories/memory_maintenance.md.
- Very specific: line 31
Use Serena memory tools when available.
- Distinctive: line 37
Read the initiative memory and its linked instructions.
- Distinctive: line 40
Use the task named by the user.
- Distinctive: line 45
Do not bypass an incomplete dependency to start a requested task.
- Distinctive: line 49
Run the task's acceptance checks.
- Very specific: line 53
Use pending, in_progress, blocked, and complete in the tracker.
- Distinctive: line 62
Use relevant available reviewers when delegation is permitted.
- Distinctive: line 69
Update the tracker, task results, and key learnings.
- Distinctive: line 72
Commit only when the current session authorizes it.
- Distinctive: line 85
Do not produce a prompt for a nonexistent next task.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Initiative
Plan work in a shared memory, then complete one task per conversation. Use the user's request to select planning or execution. Creating a plan does not start its implementation. Follow the active session's permissions and mode; in a read-only planning session, present the plan without writing memory or code.
Create a plan
- Read the project instructions, relevant code, and existing plans. Establish the intended result, scope, constraints, and measurable completion criteria. State assumptions and ask about choices that affect the work.
- Divide the work into tasks small enough to complete and verify in one conversation. Give each task its dependencies, affected files or modules, implementation steps, and acceptance checks with expected results. Follow the project's test-first requirements for code changes.
- Use assets/initiative.md to create the memory. Replace all placeholders, omit unused optional fields, and repeat the task section as needed. Retain shared context once: background, key files, design choices, and links to applicable instructions.
- Follow
.serena/memories/memory_maintenance.md. Unless the user supplies a location, useplans/<initiative-name>/initiativeas the memory name. Link it from the relevant parent memory, orplans/corefor a top-level plan, using amem:reference with a short description. Preserve existing records; do not overwrite another initiative with the same name. - Use Serena memory tools when available. Otherwise read and write the same
files under
.serena/memories/, adding.mdto the memory name. Report the memory name and provide a prompt to start the first task.
Continue a task
- Read the initiative memory and its linked instructions. Check the current branch, working tree, and relevant code before relying on recorded progress. Preserve unrelated edits and do not switch branches over existing work.
- Use the task named by the user. If none is named, resume an unfinished task already in progress; otherwise select the first pending or blocked task whose dependencies are complete. Recheck its blocker before resuming. If it still needs user input, or several tasks are already in progress, report the choice needed before editing. Do not bypass an incomplete dependency to start a requested task.
- Mark the task
in_progressand record the current agent. Implement only that task and changes needed to satisfy its acceptance criteria. If new evidence requires a scope decision, record it and ask before expanding work. - Run the task's acceptance checks. Record commands, results, and any checks that could not run. Fix failures within scope and rerun affected checks. A failed or unrun required check prevents completion.
Use pending, in_progress, blocked, and complete in the tracker. Record
the reason and the next action when blocked. Recheck that reason on resumption;
do not treat an old blocker as proof that work is still blocked. If all tasks
are complete, report the recorded results without starting additional work.
Close the task and hand off
- Review the task's diff. Select reviews that apply to the change: code correctness, error handling, tests, simplicity, comments, or type design. Use relevant available reviewers when delegation is permitted. Follow the project's test-hunter procedure for new or changed Go tests. When a reviewer is unavailable or delegation is prohibited, review directly and record that limit.
- Address findings supported by the code and within the task's scope. Record unrelated findings as follow-up work. Rerun affected acceptance checks after fixes. Unresolved findings that prevent acceptance keep the task incomplete.
- Update the tracker, task results, and key learnings. Mark the task complete only when its required checks and reviews are satisfied. Record failures and blockers accurately, including the next action needed to resume.
- Commit only when the current session authorizes it. Include only this task's changes, preserve other staged work, and follow project commit rules. Record the commit result or why no commit was made. An optional commit does not determine acceptance; if the user requires a commit for completion, keep the task incomplete until it succeeds. The skill grants no extra permission to commit, push, publish, or change external systems.
- Stop after this task's wrap-up. Do not start the next task. Present the result and a self-contained prompt for a new conversation. Include the initiative memory name, branch, completed or blocked task, next task or remaining action, and any relevant uncommitted work or verification limits. Respect an explicit user instruction to change the conversation boundary.
For the final task, report initiative completion and the acceptance results. Do not produce a prompt for a nonexistent next task. If the final task is blocked, provide a prompt to resume that task instead.
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.

