arndvs/do-work
skills/do-work/SKILL.md
Core execution loop for implementing tasks. Use when asked to 'do work', 'implement', 'build this', 'fix this', 'plan execute clear', 'plan then execute', 'work loop', or when working through a plan or backlog item.
development
Updated May 21, 2026
$ install --global
skillsauthnpx skillsauth add arndvs/ctrlshft do-workInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: May 21, 2026, 2:13 AM55.6s1 file scanned
SKILL.md
- name:
- do-work
- description:
- Core execution loop for implementing tasks. Use when asked to 'do work', 'implement', 'build this', 'fix this', 'plan execute clear', 'plan then execute', 'work loop', or when working through a plan or backlog item.
Do Work
Output "Read Do Work skill." to chat to acknowledge you read this file.
Pipeline position: /grill-me → /write-a-prd → /architect → /prd-to-issues → /do-work → shft
Workflow
0. HUD event (if daemon running)
Emit a session-start event so the HUD tracks this work:
source ~/dotfiles/bin/write-hud-state.sh
write_hud_event "info" "do-work: started — $TASK_SUMMARY"
Replace $TASK_SUMMARY with a short description. Emit again after each commit (do-work: committed — <message>) and at session end (do-work: completed).
1. Understand
Read any referenced plan, PRD, or GitHub issue. If none provided, clarify the task with the user before proceeding.
2. Plan (optional)
If the task has not already been planned, create a plan for it. Break large tasks into vertical slices (tracer bullets) — each slice should touch all layers end-to-end rather than building layer by layer.
Skip this step if a plan or PRD already exists.
3. Implement
Read a sample existing file of the same type before creating new ones — follow the conventions already established in the codebase.
For backend code: use red/green/refactor, one test at a time in a tracer-bullet style.
- Write a single failing test for the smallest vertical slice of behaviour
- Run the test — confirm it fails (red)
- Write the minimum code to make it pass (green)
- Repeat from step 1 for the next slice
- Refactor if needed while keeping tests green
Do not write all tests upfront — write one, make it pass, then move to the next.
For frontend code: implement directly without TDD.
4. Validate
Run the project's feedback loops until they pass cleanly. Auto-detect from the workspace:
- package.json → look for
test,typecheck,type-check,lintscripts. Run with the project's package manager (npm/pnpm/yarn/bun). - composer.json → look for
test,lint,phpstanscripts. Run withcomposer run. - Makefile → look for
test,lint,checktargets. Run withmake. - pyproject.toml / setup.cfg → look for pytest, mypy, ruff. Run directly.
- Pre-commit hooks → if
.husky/or.pre-commit-config.yamlexists, the commit step will trigger them automatically.
If no feedback loops are detected, tell the user and ask what validation commands to run.
5. Commit
Once validation passes, commit the work using the atomic-commits skill (one logical change per commit, conventional commit message).
6. Preflight (Ship mode only)
If this is the final slice and the work is ready for review, run the pr-preflight skill end-to-end before pushing. Preflight audits main..HEAD, so the commit must exist before it can inspect the diff. This catches the class of issues that iterative Copilot reviews find one-at-a-time, so review is one pass instead of many.
Skip this step when using Commit mode for intermediate checkpoints — preflight runs once before the final push, not on every checkpoint.
After preflight passes, use the atomic-commits skill in Ship mode — push and open a PR.
7. Context Check
If this is one phase of a multi-phase plan, or if context usage is over 40%, follow the standard handoff protocol (@~/dotfiles/instructions/handoff.instructions.md) — commit all work, persist the remaining plan to working/, and provide the pickup command.
Include @-references to research.md, PRD issues, and key files modified this session.
Related Skills
arndvs/visual-feedback
development
VerifiedTrustedCommunity
Use when implementing UI, checking dark/light mode, or validating animations — adds a visual feedback loop via browser screenshots so frontend changes are verified, not assumed.
SKILL.mdUpdated May 21, 2026
arndvs/press1-check
development
VerifiedTrustedCommunity
Use when Claude Code sessions had many manual approval ("press 1") prompts or when auditing hook permissions; identifies which Bash commands required approval.
SKILL.mdUpdated May 21, 2026
arndvs/plan-archive
tools
VerifiedTrustedCommunity
Use after merging a PR or during periodic cleanup to archive plan-mode files by linking them to merged PRs.
SKILL.mdUpdated May 21, 2026
arndvs/grill-with-docs
testing
VerifiedTrustedCommunity
Use when stress-testing a plan against the project's domain model — grills the design, sharpens terminology, and updates documentation (CONTEXT.md, ADRs) inline as decisions crystallise.
SKILL.mdUpdated May 21, 2026