Core development instructions
| ID | ap-core-instructions |
| Kind | instruction |
| Selection | required |
| Category | Core |
| Targets | claude, codex, cursor |
Project orientation, independent judgment, investigation, quality, and completion guidance.
When to use
Section titled “When to use”Always loaded after initialization. No separate chat invocation is required. Specialized workflows belong in skills and subagents so this core set stays concise.
What it covers
Section titled “What it covers”- Project orientation — read
PRD.md,TECHNICAL_REQUIREMENTS.md, andTODOs.md(or equivalents) before planning or changing code. - Independent judgment — challenge weak assumptions; surface material tradeoffs; follow explicit user direction after stating concerns.
- Clear explanations — lead with what happened, why it matters, what changed, and how it was verified.
- Investigation standard — gather evidence before concluding; verify version-sensitive APIs against current docs and the repo.
- Implementation quality — match sound conventions; prefer simple complete solutions; avoid speculative abstractions.
- Debugging and QA — treat symptoms as evidence, keep competing hypotheses, verify fixes with reproduction and checks.
- Task completion — define success criteria, run relevant checks, update docs when substantial subsystems change.
How it is selected
Section titled “How it is selected”Installed automatically with every Agents Pack initialization. It cannot be removed.
Provider delivery
Section titled “Provider delivery”Agents Pack renders this content into native locations such as:
- Claude Code rules under
.claude/rules/agents-pack/ - a managed block inside
AGENTS.md(Codex; repository or global) - Cursor rules under
.cursor/rules/agents-pack/(repository scope)
Text outside managed blocks remains yours. See Ownership and File layout.