| name | common-tones |
| description | Finds notes shared between TWO chords and describes their interval role in each (root / 3rd / 5th / 7th / extension). Used for pivot-chord choice, smooth voice leading, and modulation prep. Calls the deterministic `domain.commonTones` closure via the `ga_dsl_eval` MCP tool — never compute pivot tones from training data, since LLMs fumble extension intervals on altered chords. |
| triggers | ["common tones","common tone","shared notes","pivot tone","pivot chord","notes in common","what do these chords share","what notes do x and y share"] |
| license | internal |
| compatibility | {"agent-framework":">=1.0.0-preview","microsoft-extensions-ai":">=10.5.1"} |
| metadata | {"authoring-style":"tool-driven","origin":"graduated 2026-05-06 — second canary for ga_dsl_eval (Phase 2b)","evidence-kinds":["tool_call"]} |
| allowed-tools | ["ga_dsl_eval"] |
| last_verified | "2026-05-06T00:00:00.000Z" |
Common Tones Between Two Chords
When a user asks what notes two chords share, call ga_dsl_eval with closure name domain.commonTones. Don't compute the intersection mentally — the closure returns each shared note's interval role in BOTH chords (root / 3rd / 5th / 7th / 9th / 11th / 13th), and that role-mapping is exactly what the LLM gets wrong on extended/altered chords.
Why this skill uses ga_dsl_eval (not a keyhole tool)
This is the second canary for the DSL-eval pattern (PR #146 was the first, transpose). The domain.commonTones closure is already in GaClosureRegistry — exposed because its category is Domain. See docs/contracts/2026-05-06-ga-dsl-eval-contract.md (v0.1).
Calling the tool
Method: ga_dsl_eval
Args (flat key-value, all values as strings):
closureName = "domain.commonTones"args.chord1 — first chord symbol, e.g. "Cmaj7", "F#m7".args.chord2 — second chord symbol.
Example invocation:
ga_dsl_eval(closureName: "domain.commonTones",
args: { "chord1": "Cmaj7", "chord2": "Am7" })
Returns a DslEvalResult. Read Result for the formatted shared-notes string (or ResultJson for the same value as a JSON-quoted string). On failure, read Error.Code and Error.Message.
Mapping user phrasings
- "What notes do Cmaj7 and Am7 share?" →
args = { "chord1": "Cmaj7", "chord2": "Am7" }
- "Common tones between G and D7" →
args = { "chord1": "G", "chord2": "D7" }
- "Pivot tones from F to Bb7" →
args = { "chord1": "F", "chord2": "Bb7" }
- "What's shared between Cmaj9 and Em7?" →
args = { "chord1": "Cmaj9", "chord2": "Em7" }
Phrasing the answer
The closure returns a formatted string already. Surface it verbatim, then add a one-line interpretive hook tailored to the user's apparent intent (pivot, voice leading, modulation prep). Example:
Cmaj7 and Am7 share these notes:
- C → root of Cmaj7, minor third of Am7
- E → major third of Cmaj7, perfect fifth of Am7
- G → perfect fifth of Cmaj7, minor seventh of Am7
Three common tones — that's why Am7 is the canonical relative-vi substitute for C. Swap them anywhere in a progression with minimal harmonic disturbance.
If the closure returns "X and Y share no common tones", surface that and explain what it means: harmonically distant chords (often a tritone-related pair, an altered-dominant against a parallel major, or genuinely opposing modal centres). Suggest checking interval distance via the interval skill or using chord-substitution to find a related pair.
When to refuse / clarify
- Single chord — that's a
chord-info question (notes of one chord). Defer.
- Three or more chords — out of scope for v0.1 (closure takes pairs). Either compute pairwise (Cmaj7-Am7, Am7-F, Cmaj7-F) or ask the user which pair matters most.
- Unparseable chord symbol — the closure returns a
ParseError naming which chord (chord1 or chord2) failed; surface the error and ask for the correct symbol.
Out of scope
- N-way common tones (N ≥ 3) — closure takes pairs only. Wrap with multiple calls if you need a 3-way intersection (compute pairwise, then intersect the result sets in prose).
- Voice-leading optimisation — separate skill (BACKLOG #139 gap, not yet shipped). Common tones are a sub-step of voice leading; this skill returns the intersection, not the optimal motion of all voices.
- Modulation suggestions — knowing the common tones is foundational, but choosing a destination key for modulation is a richer query. Defer to
key-identification for analysing existing progressions.
Cross-reference
- MCP tool surface:
ga_dsl_eval in Common/GA.Business.ML/Agents/Mcp/DslEvalMcpTools.cs
- Closure:
domain.commonTones in Common/GA.Business.DSL/Closures/BuiltinClosures/DomainClosures.fs (line ~489)
- Contract:
docs/contracts/2026-05-06-ga-dsl-eval-contract.md (v0.1)
- Plan:
docs/plans/2026-05-06-skills-orchestration-architecture.md (Phase 2b — second canary)
- Sibling:
transpose (first DSL-eval canary, PR #146)
