task-implementation
Skill2 starsChanged 22 days agoMITView on GitHub
Execute an approved task; implement within the single write scope, compare expected against actual verification, record deviations and recover from checkpoints. Does not auto-escalate design or review requests into implementation.
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
Verify the task is not yet done and its dependency artefacts and interface versions are satisfied.
- Very specific: line 26
Use the implementation receipt to give actual changes, expected/actual, unverified items, deviations and the recovery entry.
≈ “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
Implementing per approved task
Turn an existing approval into verifiable file changes and results. A simple change the user states explicitly can itself constitute the task and its authorisation; do not build a full plan to satisfy a format. When a plan exists, consume its exact version, acceptance and dependencies instead of redesigning one.
Start or resume
Read the current goal/mode, approved scope, tasks and dependencies, writable files, sole writer, existing checkpoints and relevant project commands. Missing approval blocks only implementation; read-only clarification may proceed. Missing write permission means that file is not touched. When the plan still has open questions, separate those that block this task from work that can be completed independently.
On resume, compare the checkpoint against the actual candidate files, run evidence and related execution state. Parts marked done whose content or dependencies changed need re-verification; parts truly done and still applicable are not redone. While it is unknown whether an old execution has stopped, stay read-only and verify in-flight work through subtask-dispatch or the existing execution-status flow; never give the same file a second writer.
Loop per task
- Verify the task is not yet done and its dependency artefacts and interface versions are satisfied. Read the task's real inputs and existing implementation; do not guess function signatures, field semantics or command parameters from earlier summaries.
- Define observable expectations: what the user will see, how errors are rejected, how boundaries or state recovery behave; reduce the gap between plan requirement and current state to the minimal change. Ordinary implementation details follow existing project patterns at your discretion.
- Implement within the authorised scope. A regression case for a bug fix or new behaviour should first be observed failing, then pass after the fix; the failure must come from the target behaviour, not a missing dependency or syntax error. For pure copy/style/generated artefacts choose a meaningful check; do not build a mechanical TDD ritual.
- Run the relevant lint, build, tests and necessary manual checks the project prescribes, read the results and compare each against expectations. Save candidate identity, environment/inputs, full log location and key assertions. Exit 0 is only the process result; wrong data, a failed business assertion or a skipped case is still not a pass.
- When expectations are not met, use Deviation handling to decide whether it is an implementation error, a verification problem, a plan conflict or a missing environment item; fix the cause, then re-verify affected items. Necessary checks and fixes within scope continue to completion; do not re-request already given authorisation at every step.
- Reclaim containers, images, processes, temporary files and test data produced this run per Local resource cleanup: register on creation, delete only registered own objects at wrap-up, and list retained evidence and cleanup failures in the receipt.
- Use the implementation receipt to give actual changes, expected/actual, unverified items, deviations and the recovery entry. Hand to
change-review/verification-gateor the project's existing review flow; author self-tests are not independent acceptance, and one passing log does not declare everything finished.
Completion and handover
Before declaring completion confirm: artefacts within the approved scope are complete, necessary checks actually ran with matching results, deviations are explained, temporary resources from this run are reclaimed or remaining items have stated reasons, and unfinished items have accurate impact and next step. Checks that cannot run are marked unverified; if such a check is an acceptance prerequisite, the task still awaits evidence. Whether old verification still applies after a change is judged by impact; never reuse a report that does not match the current candidate.
Hand only this task's necessary facts to the existing status maintainer; the current user workbench/checkpoint is carried by context-handoff. This package keeps no second progress store. After stopping writes, state the write handover and remaining in-flight work. Without the companion packages, use the project's existing plan, status and verification gate.
Resources and examples
Ordinary implementation starts the estimate at normal/high; low/low suits only mechanical items whose inputs, outputs and change location are precisely fixed. Grade words map to actual execution configuration through the project resource mapping. When cross-interface complexity rises, re-estimate dependencies and risks first; use subtask-dispatch only when independent work is genuinely needed and the host allows it. Do not bypass approval by switching model or tool.
Normal example: an approved fix for empty-list pagination first reproduces the out-of-range error with an empty list, then changes the boundary handling, runs the necessary checks and gives evidence tied to the candidate. Counterexample: refactoring another module without write permission "to make testing easier" must stop and report the boundary. Resume example: the checkpoint has data-layer changes and valid tests but the UI is unfinished; after verifying the candidate matches, complete only the UI and the affected integration checks. Failure example: the command exits 0 but the returned record count does not match acceptance; record the failure and keep investigating, do not fill in pass.
Sources
- Pinned source: SP04 executing-plans.
- Adopted the per-task loop, real inputs, expected-versus-actual comparison and completion evidence; approval boundaries, reuse of existing checkpoints, deviation classification and single-writer recovery are own-authored for this package. Dropped automatic Git/worktree and upstream scripts, mechanical TDD, self-adjudication of arbitrary conflicts and a default highest-cost final review.
- License: shipped with the package as LICENSE-SP.txt (SP, MIT); library-wide third-party summary in the root THIRD_PARTY_NOTICES.md. Copy the license file along when copying this package alone.
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- ItsSidhartha/claude-skills★ 0same name
- mustafahasanain/frappe-workflow★ 0same 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 repository39 other files it gives its agents.
- ai-trace-auditSkill
- ai-workspace-governanceSkill
- archive-maintenanceSkill
- change-reviewSkill
- cloud-api-integrationSkill
- coding-conventionsSkill
- container-image-managementSkill
- context-handoffSkill
- deployment-patternsSkill
- docker-local-environmentSkill
- documentation-authoringSkill
- estimationSkill
- external-dependency-simulationSkill
- gitlab-webhook-local-gitSkill
- go-patternsSkill
- go-testingSkill
- grpc-proto-contractSkill
- guided-evidence-collectionSkill
- http-api-designSkill
- jenkins-pipeline-jjbSkill
- ldap-rbacSkill
- mysql-migration-safetySkill
- plan-designSkill
- plan-reviewSkill
- playwright-e2eSkill
- project-contextSkill
- pytest-patternsSkill
- readonly-ops-diagnosticsSkill
- redis-patternsSkill
- release-readinessSkill
- role-bootstrapSkill
- security-reviewSkill
- skill-maintenanceSkill
- subtask-dispatchSkill
- systematic-diagnosisSkill
- test-strategySkill
- verification-gateSkill
- vue3-developmentSkill
- vue-testingSkill
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.

