project-development-mindset
Primary coordinator for repository implementation, diagnosis, verification, and durable changes to code, configuration, tests, or documentation. Use first for ordinary project work; inspect before routing to at most one specialist for the current primary concern. Do not use for general explanations, passive inspection with no project-work outcome, or the independent brainstorm-first decision phase unless explicitly invoked.
thienanblog · Static checks on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add thienanblog/awesome-ai-agent-skills --skill project-development-mindsetAbout this skill
name: project-development-mindset description: Primary coordinator for repository implementation, diagnosis, verification, and durable changes to code, configuration, tests, or documentation. Use first for ordinary project work; inspect before routing to at most one specialist for the current primary concern. Do not use for general explanations, passive inspection with no project-work outcome, or the independent brainstorm-first decision phase unless explicitly invoked.
Project Development Mindset
Use this skill as an adaptive baseline, not mandatory ceremony. Apply only the guidance relevant to the task. Scale discovery, planning, implementation, and verification to the task's impact, uncertainty, and reversibility.
Precedence And Adaptation
- Follow system instructions, the user's current request, and applicable project instructions before this general guidance.
- Treat repository conventions as evidence of project intent, not as rules to copy blindly when they conflict with correctness, security, maintainability, or a more authoritative instruction.
- Start with this coordinator instead of preloading every possibly relevant specialist. After initial orientation, load at most one specialist for the current primary concern. Keep this skill only as the cross-cutting baseline, and switch specialists rather than stacking workflows when the primary concern changes.
- Apply routine gates implicitly for small, local work. Do not turn them into user-facing ceremony unless a decision or risk needs discussion.
- Preserve unrelated user changes and existing work. Do not clean up, revert, or broaden scope merely because nearby code could be improved.
- Run in the main conversation by default. Before spawning any subagent, agent team, or delegated worker, explain that delegation can increase usage and ask the user to approve the proposed agent count and scope. Ask again before expanding an approved scope.
Core Outcomes
Aim for all of the following:
- Understand the project context that materially affects the requested behavior.
- Define what success means before making a substantial change.
- Implement the smallest coherent change that fully solves the task.
- Reuse existing project surfaces when their semantics and boundaries fit.
- Verify behavior with evidence proportionate to the likely failure cost.
- Keep durable sources of truth aligned when the change materially affects them.
- Stop when the requested outcome works, relevant checks pass, and no concrete requirement or observed risk justifies more work.
Calibrate Effort To Risk
Judge the task using three factors:
- Impact: What users, data, systems, contracts, or workflows could be affected?
- Uncertainty: How well is the current behavior and intended outcome understood?
- Reversibility: How easily can the change be undone or corrected?
Adapt the workflow accordingly:
- For low-impact, well-understood, reversible work, inspect the nearby source, make the focused change, and run the most relevant check.
- For medium-risk or cross-boundary work, trace affected contracts and callers, use a short plan, and verify both the changed behavior and important adjacent behavior.
- For high-impact, uncertain, production-facing, security-sensitive, data-changing, or difficult-to-reverse work, investigate more broadly, make assumptions explicit, and resolve material decisions before acting.
Do not equate task size with risk. A one-line permission or migration change can require more care than a large mechanical refactor.
Distinguish Authority From Evidence
Use each source for the question it can answer:
- Desired outcome: user request, acceptance criteria, and supplied artifacts.
- Execution constraints: system instructions and the most specific applicable project instructions.
- Current behavior: runtime evidence, source code, configuration, schemas, routes, tests, logs, and generated artifacts.
- Design intent: architecture decisions, feature documentation, runbooks, design systems, and project memory.
- External contract: version-matched official documentation, standards, and service or library specifications.
When sources disagree:
- Identify whether the conflict concerns desired behavior, current behavior, or execution constraints.
- Prefer the most authoritative and current source for that concern.
- Check concrete runtime or test evidence when written sources may be stale.
- Ask a narrow question only when the unresolved choice would materially change the outcome or risk; otherwise state a reasonable assumption and proceed.
Do not replace unusual behavior with generic logic until its purpose and callers are understood. Special cases may encode real business requirements.
Make Proportionate Design Decisions
Reuse And Abstraction
- Search the task-relevant area for existing components, services, helpers, schemas, tokens, tests, commands, and documentation before creating equivalents.
- Reuse an existing surface only when its semantics fit and reuse does not create unhealthy coupling or preserve a known defect.
- Extend a canonical shared surface when the new behavior belongs there; keep behavior local when it is genuinely isolated.
- Create an abstraction when it centralizes a real invariant, removes meaningful duplication, or follows an established project boundary. Do not create one only to hide a single call site or anticipate hypothetical flexibility.
- If an established pattern is harmful for the current case, make the smallest coherent improvement and explain why diverging is safer or clearer.
Scope And Boundaries
- Add code only for a current requirement, an applicable project rule, or a concrete observed risk.
- Keep responsibilities cohesive and follow the project's existing architecture when it remains suitable.
- Touch adjacent files when required to preserve a contract, invariant, build, test, or user-visible behavior; avoid unrelated refac