rosetta[scheme]: cross-language consistency — 5 red / 3 amber metrics, 5 ledgered shapes

[squid-protocol]

Part of #2560 — per-language tracking for the keyword-rosetta control corpus (built for #1096). Planted intent is identical in every language, so every deviation below is measured bias vs. the 46-language median, validated in the corpus's deviation ledger. 🔴 = beyond ±50% of the cross-language median, 🟡 = ±25–50%.

Triage order (per the epic): structure detection first, then signal/keyword rules, then risk-score consequences — §3's numbers are downstream of §1–§2 and re-baseline as those fix. Cross-cutting engine issues (e.g. #2535/#2546) fix many languages at once; this issue tracks this language's residual after each cross-cutting fix lands, and closes when the language sits inside the ±25% band on every comparable metric (or every remaining deviation is ledgered as intended morphology).

1. Structure detection (fix first)

• 🔴 dependency_links: 0 vs median 3 (-100%)

2. Signal / keyword deviations

• 🔴 planned_debt: 2 vs median 1 (+100%)
• 🟡 branch: 3 vs median 5 (-40%)
• 🟡 high_risk_execution: 3 vs median 2 (+50%)
• 🟡 state_mutation: 3 vs median 6 (-50%)

3. Risk-exposure consequences (downstream)

• 🔴 risk_api_exposure: 0 vs median 5.31104 (-100%)
• 🔴 risk_cognitive_load: 4 vs median 15.1444 (-74%)
• 🔴 risk_tech_debt: 80 vs median 39.9576 (+100%)

Ledgered shapes touching scheme

defect	type	issue	evidence	summary
census-requires-git-tracked	engine-semantic	—	—	GalaxyScope's census enumerates git-tracked files only; an untracked folder scans as '0 files mapped' with no per-file warning
batch4-dual-keyword-overlaps	keyword-overlap	—	—	Collective entry for batch 4's finditer-confirmed dual-rule keywords, each detailed in its language's manifest notes: agc RESUME (cleanup+branch), livecode global (globa…
cobol-hyphen-identifier-debt-leak	upstream-bug	#2537	—	Debt rules are whole-text, not comment-anchored, and '-' is a regex word boundary: HACK-LEVEL (data item, 2 code occurrences) counts fragile_debt, PROBE-TODO (paragraph …
no-dependency-capture-languages	upstream-question	—	—	m4 and scheme have _dependency_capture = None: include()/(import ...) count the import signal but can never produce a DAG edge, so popularity stays 0 and orphans never c…
string-literal-selective-shielding	upstream-question	#2535	—	Keywords inside one string literal ('if eval fails, try open') count toward branch (+2), safety (+1), io (+1) but eval does NOT count toward high_risk_execution (routes …

Decoy outcomes (file-level evidence)

• b.scm · (set! note "if eval fails, try open") (string): scheme's branch keywords are sexp-anchored — bare 'if' inside the string does NOT count branch (syntax shielding) — but bare eval DOES count high_risk +1 (its rule is not anchored). state stayed exactly 3 raw, no flux (no branch context). Shielding by anchoring, a fourth profile.
• c.scm · (define (probe-todo plan) ...) (code): COUNTED: hyphen is a word boundary so \bTODO\b matches inside probe-todo — planned_debt 2 (comment + identifier). Scheme joins cobol in gitgalaxy#2537's hyphenated-identifier debt leak.

Reproduce / evidence

• Corpus folder: data/scheme/ — expected_signals.json carries the shape-by-shape accounting
• Gate: GALAXYSCOPE_BIN=<venv>/bin/galaxyscope python tools/verify_language.py scheme in the corpus repo
• Full context: findings_by_language.md#scheme · bias_report.md · bias chart