fix(embedded_python): state_mutation hardware arm counted getter reads as writes

[squid-protocol]

Problem

embedded_python's state_mutation hardware-toggle arm \.(?:value|on|off|high|low|toggle)\s*\( matched on call syntax alone, so it counted reads as writes — a violation of the state_mutation contract (#2765, docs/state_mutation_rule_contract.md):

• Corollary 3 (a read is not a write): if recovery_pin.value() == 0: (a getter) counted as a mutation.
• Corollary 4 (a token another rule owns is not a second signal): val = self.pin.value() double-counted — the assignment arm already counts the write to val, and the .value() getter added a spurious second hit; and a parameterised .on(evt, cb) is an event subscription that events/listeners own.

This is not a contract change — state_mutation is already stated (#2765). It's a single-language regex fix under that contract (per the harden-language-extraction skill, which explicitly excludes rule-contract-audit for single-language regex work), where a corpus measurement showing the rule firing on a read "is a bug in the rule, not a quirk of the corpus."

Fix

.value( is both getter and setter in MicroPython/CircuitPython, so it now requires a non-empty argument (pin.value(1) writes, pin.value() reads). .on/.off/.high/.low/.toggle are the no-arg imperative write forms and now match the no-arg shape only, so .on(evt, cb) no longer reads as a mutation. Only tightens quantifiers → no ReDoS surface added.

Verification (investigated against the real meow_turtle corpus, 14 files)

• Every real write preserved: all corpus .on()/.off()/.toggle() are no-arg (kept); the 5 .value(arg) setters still count. Only the 2 getter reads drop (raw 894 → 892).
• test_state_mutation_contract_2765.py: added getter/subscribe negatives + one-statement-one-hit counts (val = self.pin.value() → 1, if pin.value() == 0: → 0, led.on() → 1). 102 pass.
• Golden masters regenerated (both legs); bless_scope = 15 substantive + 5 topological diffs, all embedded_python/meow_turtle (mutation signal, its weighted/cascading views, downstream mass/exposure rollups), no file newly parsed or excluded.
• rosetta 46/46, 0 regressions (embedded_python fixture uses none of these tokens) — no paired corpus rebless owed.
• ruff_audit/mypy_audit/dead_key_audit --ci: no new findings. Full pytest tests/: 8746 passed (only the pre-existing test_security_auditor.py xgboost-env failures, identical on origin/main).

Contract audit-table row for embedded_python updated (191 -> 894 -> 892) with the corollary 3/4 rationale.

🤖 Generated with Claude Code

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab