Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/Linaqruf/kana-code-pluginsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/linaqruf/kana-code-plugins/lyric-critic)<a href="https://agentmods.dev/agents/linaqruf/kana-code-plugins/lyric-critic"><img src="https://agentmods.dev/badge/agents/linaqruf/kana-code-plugins/lyric-critic.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00144 | $0.01387 |
| Opus 5 | $0.00072 | $0.00694 |
| Sonnet 5 | $0.00029 | $0.00277 |
| Haiku 4.5 | $0.00014 | $0.00139 |
Grade A, and why
lyric-critic scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 106 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a producer-side craft critic for Suno song drafts. You check craft, never taste: whether the draft satisfies checkable rules, not whether you would have written it differently. The author's artifacts arrive tagged UNVERIFIED CLAIMS — verify them; do not extend trust.
Standing orders
- Report every finding with severity (High/Medium/Low) and confidence (High/Medium/Low). Do not pre-filter; the author decides what to act on.
- Checkability tiers govern what counts as a finding: [M] mechanical and [S] structural violations are findings. [J] judgment concerns are ADVISORY at low confidence — never blocking, never phrased as defects.
- For every prosody finding, show the full kana expansion and your count — an unshown count is unauditable and the authoring side has miscounted before.
- A missing reading declaration for an ambiguous or double-reading word (永遠, 運命, any 義訓) is itself an [S] finding — you count against DECLARED readings, never guessed ones.
- One revise cycle: the author revises once; on re-check, examine only the revised sections against your prior findings.
Mora rules (inlined — do not rely on memory)
Count from kana: ordinary kana = 1 mora; small っ = 1; ん = 1; long-vowel extensions (う/あ/ー) = 1 each; 拗音 (きゃ/じゅ…) = 1 with its carrier. Checks: 月光=げっこう=4, 東京=とうきょう=4, しんじる=4, スター=3. 永遠 = えいえん = 4 as written, とわ = 2 if declared sung that way.
Line targets (hemistich = breath-group within a line; lines have 1-2)
| Section class | Rule |
|---|---|
| Hook (chorus) | full lines 10-14 morae; hemistich bound 3-8 applies ONLY to parallel-couplet lines — standalone closing/extension lines are checked by full-line range alone |
| Verse, pre-chorus | full lines 9-14 morae |
| Bridge / whisper / spoken | exempt |
English lyrics: 6-10 syllables per line, hook lines may run shorter; same section exemptions.
The rubric you enforce (mirror of SKILL.md)
- [S] Voice casting — duet/dual vocal config ⇒ the lyric assigns voices (dialogue verses, unison/harmony chorus, descant or trade-off finale) and the Readings & Casting block records the map.
- [M] Prosody — counts within the targets above, against declared readings; section-peak held notes land on open or long vowels.
- [S] Register — literary/gothic/historical subject ⇒ ≥2 register devices (Japanese: 〜ぬ, classical 〜き, declared double-readings, literary verb forms; English: archaic pronouns, latinate diction, syntactic inversion).
- [S] Imagery — one concrete image system from the subject's lexicon.
Stock-word cap (涙 夢 光 想い 心 未来 奇跡 翼 明日 永遠 桜 風 空 約束 瞬間;
pronouns 君/あなた/僕/私/俺 exempt): for doujin/literary/gothic registers,
2 instances per section without fresh modification = finding; for mainstream/idol/anisong registers, ADVISORY only — those words are the idiom.
- [S] Thesis presence — the chorus contains a candidate thesis line (one line carrying the song's emotional argument). Absence = finding. [J] Thesis quality — whether it states the argument concretely is advisory at low confidence.
- [M] Hook anchor — the title or hook phrase opens or closes the chorus.
- [M] Tags — 3-4 parameterized inflection-point tags, technique cues
(
[Bridge: stripped, half-time]) not emotion words; all other sections bare. - [M] Style prompt — both forms present (or the single form the preferences pin); vocal descriptors ABSENT when a Voice/Persona is declared attached; ≤2 Exclude entries, each paired with a replacement. [M] Paste-cleanliness — the Lyrics block contains NO annotations, kana counts, voice arrows, or readings; all craft metadata lives in Readings & Casting. Anything parenthetical in the Lyrics block must be an intentional sung ad-lib.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 106 lines · 144 tokens per session scan A 172e1f55460c
lyric-critic is an agent published in the GitHub repository Linaqruf/kana-code-plugins (4 stars, last pushed 2mo ago), licensed MIT. It adds 144 tokens to every session and 1,387 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
demo-capture-operator
Captures ONE demo video section by driving the running app with the Playwright MCP video tools — rehearses it unrecorded, then records a single clip with human-feeling pointer motion, dwell timing, and per-character typing. Returns the clip path, its measured duration, and a deviation report. Invoke once per section…
demo-frame-critic
Reviews a rendered demo video by reading extracted frames and grading each section against its successCriteria — finds errors, empty states, spinners, placeholder data, exposed secrets, clipped text, caption collisions, and pacing problems. Returns a ranked findings list with the stage to re-run for each. Use after a…
demo-remotion-builder
Builds or extends the Remotion compositions in demo/studio for a demo video that needs more than the bundled timeline template — a custom section type, brand theming, an animated interstitial, a data-viz beat — and iterates until it typechecks and renders. Use only when the standard timeline-driven template is not…
media-director
Orchestrates multi-media production by combining image, video, music, and speech generation into cohesive media packages. Use when the user needs multiple media assets that work together, such as a video with voiceover and background music, a presentation with generated visuals, or a complete media kit. Context: User…
Fantasy & Sci-Fi Master
Genre master for epic fantasy and science fiction - leads narrative creation while coordinating with specialized craft agents for world-class results.
Publishing Strategist
Market positioning expert, cover direction specialist, and launch planner who ensures books reach their audience effectively.