CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

codex-task-waves

Use when the user says "have codex fix this" / "have codex implement this" / "let codex handle this" / "give this to codex" / "delegate this to codex" for a single task with context already in scope (a Jira ticket, GitHub issue, file diff, bug, or described change). Plans the work, splits it into reviewable waves, dispatches codex per wave with review and correction between waves before opening a PR. Not for multi-issue parallel batches (use codex-issue-waves) or one-shot codex runs without planning (use invoking-codex-exec).

Quality

95%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

88%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A highly actionable orchestration manual: every step carries concrete commands, verbatim prompt contracts, explicit decision branches, and escalation caps, with strong validation and feedback loops. The only costs are mild redundancy between the dedicated sections and the red-flags table, and the fully inlined phase-logging subsystem that would sit better in a reference file.

Suggestions

Move the WAVE_PHASE logging details (line format, the wave_log shell function, and usage examples) into a references/ file and keep a 3–4 line pointer in SKILL.md, slimming the core orchestration flow.

Deduplicate the red-flags table against the dedicated sections — for rows that restate 'Read-only enforcement' and 'When the JSON is malformed' procedures, keep the symptom and fix columns terse and point to the section rather than restating the full recovery steps.

Trim or compress the 'Why codex for review' rationale section to one line, since the boundary is already enforced by the 'Review delegation — codex only' heading and the red-flags table.

DimensionReasoningScore

Conciseness

The body is dense and operational with no basic-concept padding (no "what a worktree is" explanations), but is mildly redundant: boundary-violation handling and malformed-JSON recovery each appear both in dedicated sections and again in the red-flags table, and "Why codex for review" is rationale rather than instruction. This fits 4 ("minor instances of over-explanation that could be trimmed") better than 5 ("every token earns its place").

4 / 5

Actionability

Fully executable guidance throughout: concrete commands (`git status --short`, `git diff <last-wave-commit>..HEAD`, `git rev-parse HEAD` + `git status --porcelain`, `python3 -c "import json,sys; …"`, `gh pr create`), a copy-paste-ready `wave_log` bash function, verbatim recovery prefixes ("PRIOR REVIEW OUTPUT FAILED TO PARSE AS JSON. Output ONLY a single valid JSON object…"), and exact per-dispatch prompt specifications. Edge cases (malformed JSON, boundary violations, non-convergence) all have concrete prescriptions.

5 / 5

Workflow Clarity

Clear sequencing (Phases 1–5 plus a 7-step per-wave loop) with explicit validation checkpoints (status verification, full project verification per wave, "Move to the next wave only when the current wave is clean"), explicit feedback loops (corrector → re-review loop capped at 3 cycles before escalation; re-dispatch-once for malformed JSON), and checklists (verbatim reviewer checklist, red-flags table). Validation is pervasive, so the destructive/batch cap does not apply.

5 / 5

Progressive Disclosure

Sub-skill references are one level deep and clearly signaled with named sections ("See `invoking-codex-exec` 'Codex roles' and 'Orchestrator boundaries'") and headers make navigation easy, but no bundle files exist — the entire ~200-line operational manual is inlined, including the ~45-line WAVE_PHASE logging subsystem whose script and usage examples belong in a reference file. This is "good structure; minor organization gaps" (4) rather than 5, since the under-50-line no-references exception does not apply.

4 / 5

Total

18

/

20

Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

An exemplary description: explicit third-person action list, five verbatim natural-language triggers, and explicit negative boundaries against the two neighboring codex skills. Both what and when are answered concretely with no filler.

DimensionReasoningScore

Specificity

"Plans the work, splits it into reviewable waves, dispatches codex per wave with review and correction between waves before opening a PR" lists multiple concrete pipeline actions with comprehensive coverage of the workflow, matching the 5 anchor; it clearly exceeds 4 ("several specific actions; minor gaps") since no pipeline stage is missing.

5 / 5

Completeness

Both halves are explicit: an explicit "Use when the user says…" trigger clause with concrete phrases, and a concrete "what" (plan → split into waves → dispatch codex per wave with review and correction → open PR). This matches the 5 anchor example pattern exactly; the mandatory-trigger guideline is satisfied.

5 / 5

Trigger Term Quality

Five natural user phrasings are quoted verbatim — "have codex fix this" / "have codex implement this" / "let codex handle this" / "give this to codex" / "delegate this to codex" — plus context types ("Jira ticket, GitHub issue, file diff, bug, or described change"), giving comprehensive synonym coverage a user would actually say.

5 / 5

Distinctiveness Conflict Risk

"Not for multi-issue parallel batches (use codex-issue-waves) or one-shot codex runs without planning (use invoking-codex-exec)" explicitly disambiguates the two nearest sibling skills, and the trigger phrases ("have codex fix this" etc.) define a clear single-task niche with minimal conflict risk.

5 / 5

Total

20

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

ddnetters/homelab-agent-skills

Path

codex-task-waves/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.