fix(mcp): declare Frida status session mutation - #3902

Summary

Fixes #3326. Match Frida status effect metadata to its existing Evidence recording.

 frida_session_status
- readOnlyHint: true,  mutatesSession: false
+ readOnlyHint: false, mutatesSession: true
  destructiveHint: false

Problem and expected behavior

Successful status queries retain Evidence, but the catalog advertised them as
read-only. REA's read-only definition includes session state. Ordinary additive
Evidence recording should remain non-destructive.

Change and scope

Reuse the existing evidence effect declaration in toolEffects.ts. Extend the
existing MCP lifecycle regression; no handler, Frida engine, target behavior,
runtime permission, or lifecycle change. Advance maintainer scope decision is not
applicable to this ordinary metadata correction.

Contract and boundary impact

• Semantic owner: canonical tool effects, before annotation/inventory projection.
• CLI/MCP: correct tools/list and binary_session metadata; input/output shapes
  and the one-shot CLI are unchanged.
• Provider/platform: unchanged; no new native Frida compatibility claim.
• Evidence: status output and retained Evidence remain unchanged.
• Target selection, execution/mutation authority and cleanup: unchanged.
• Generated metadata: catalog projections regenerated/checked locally; ignored
  generated outputs are not included. No version, package or installation edit.

Expected effect of the corrected status operation: additive in-memory session
Evidence recording. It does not introduce target execution, instrumentation,
target/file/configuration mutation, network access, or protection recovery.

Evidence and regression coverage

• Before: on baseline 788964d52, the new MCP regression failed at
  readOnlyHint: true versus expected false (1 failed / 25 passed).
• After: the status response is schema-valid and retained unchanged;
  tools/list advertises non-read-only/non-destructive behavior and binary_session
  declares mutatesSession. An unknown session adds no Evidence.
• Production manager/server/ledger, with a simulated external Frida device on
  Windows. This proves the metadata/retention boundary, not real-engine behavior.

Validation performed

Windows x64; pinned Node 24.18.0 / npm 11.16.0:

• npm ci — passed; 0 vulnerabilities reported.
• npm run test:focused -- src/contracts/toolEffects.test.ts src/contracts/fridaToolContracts.test.ts tests/boundary/mcp/fridaMcp.test.ts tests/boundary/mcp/fridaLifecycleMcp.test.ts
  — 40 passed, 0 skipped.
• npm run check — passed, including build dependencies, typecheck, lint,
  format, knip, metadata, module boundaries and architecture guards. Warnings remain.
• npm run docs:check — passed, including catalog generation/checks and 5 README
  consistency tests.
• git diff --check — passed.
• Compiled production stdio MCP — tools/list reports readOnlyHint false and
  destructiveHint false; binary_session advertises mutatesSession, without an
  injected provider. No real Frida target was selected or executed.

Compatibility, safety, and release

No payload/schema-shape change or migration. MCP clients may now seek confirmation
for this correctly classified operation, depending on their own policies.
No real Frida/device lane or full-suite run; native implementation is unchanged.

Merge Danger

Door: two-way. Blast radius: narrow — one tool's effect classification and the
derived catalog identity; no persistent-data migration or native behavior change.

Author readiness

Human owner: @z2Ace0107

• I reviewed the final diff and relevant existing code, understand the change and its edge cases, and take responsibility for the complete PR, including AI-assisted work.
• I checked the evidence above, confirmed it supports the current change, and recorded meaningful validation gaps.
• I declare this PR ready for initial maintainer review.

Before merge

• This PR addresses a concrete problem or an agreed enhancement.
• Material scope and authority questions are resolved, and any required maintainer scope decision covers the final diff.
• The PR has one focused outcome and the title follows type(scope): outcome.
• Related issue is linked, or the reason for not linking one is stated above.
• Tests cover changed observable behavior and meaningful failure paths.
• Owning docs, contracts, and generated metadata are updated where needed.
• User-visible CLI/MCP changes include representative output.
• I checked the final diff for secrets, unrelated cleanup, and unsupported claims.