Agent skills for Codex CLI

This skill has no YAML frontmatter at all.

For "No High Risk Confirmed" — the end-to-end roadmap: outlines standards, burden of proof, and counterarguments; result: test product…

Enter "offline mode" — for security reasons, do NOT use any tool or command that touches the internet or an external network.

Use when reporting findings, progress, or asking for approval to someone who owns the product but does not read the code, when the reader…

Use when product or application logic may contain contradictory states, mixed date or filter scopes, KPI/card mismatches, misleading…

Use when reviewing a product artifact or implementation handoff that contains proposed or implied screens, navigation, roles, workflows…

Use when reporting task status, completion, blockers, percentages, ETAs, test results, deployment readiness, or handoff progress where…

Use when fixing bugs, refactoring, changing shared state, or reviewing a patch where behavior that already works could regress; triggers…

Use when planning, implementing, debugging, or reviewing work where the solution could expand beyond the user's requested outcome…

Silent mode — do the work with NO conversational talk.

Never burn time blocking on a wait. Forbids foreground `sleep`/idle-polling to wait for a long job (a download, fetch, build, backtest…

Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in.

Spawn N parallel candidates at the same task, pick a base, graft the strongest parts of the losers into it.

Design an auditable playbook when no narrower one fits: a large migration, an ambitious multi-part change, or work a human reviews after…

Use for \"how does X work\", code walkthroughs before changing something, and placement / ownership / layering questions (\"where should…

Index of the engineering principles. Load it whenever work is non-trivial: designing types or a signature, refactoring, debugging…

Reconstruct your recent working context from your own chat history, live state, and the shared record (user reports, prior fixes…

Keep a reviewable decision trail for long-running or unattended work: a TSV log with one row per decision (what, why, evidence, result).

Fan out N parallel workers, drain them, and return one report.

Process large datasets efficiently using chunk(), chunkById(), lazy(), and cursor() to reduce memory consumption and improve performance

Portable storage configuration across S3/R2/MinIO with optional CDN—env toggles, path-style endpoints, and URL generation