52 lines (37 loc) · 1.9 KB

Contributing

Thanks for helping improve Codex Sentinel.

Good Contributions

clearer stage detection rules
better stack-specific guidance for supported ecosystems
new stack profiles with concrete framework-specific red flags
better example prompts and adoption docs
tighter wording that reduces false confidence

Ground Rules

keep each SKILL.md concise
keep description frontmatter focused on triggering conditions, not workflow summaries
move repeated details into shared references instead of duplicating them across skills
do not add language that claims a repository is secure or fully reviewed
do not add instructions that silently install tools or mutate project files unless that behavior is explicitly scoped and documented

If You Add A New Stack Profile

Add a file under skills/shared/stack-profiles/.
Keep the same section shape:
- When To Use
- Focus Areas
- Red Flags
Update any example prompts or docs that should mention the new stack.

Validation Checklist

Before opening a PR, verify:

every skill has name and description frontmatter
shared references stay the source of truth for common rules
prompts and docs clearly distinguish planning gaps from confirmed findings
no new content creates false assurance

Workflow Dependencies

this repository owns the suite's activation rules, guidance text, examples, and validation contract
upstream workflow skills can still affect the end-to-end user journey in live Codex sessions
when validation fails, distinguish a local suite defect from upstream dependency drift before proposing a fix
do not paper over upstream workflow gaps by making local docs promise behavior the suite cannot guarantee directly

Review Notes

When proposing changes, explain:

which stage the change affects
whether it changes opt-in behavior
whether it changes shared reporting or stack-specific logic

⚡