feat(agent): record coding verification evidence

[OutThisLife]

Summary

• add a profile-scoped verification evidence ledger for canonical project checks detected by agent.coding_context
• record foreground terminal test/lint/typecheck/build results as scoped evidence (full vs targeted, pass vs fail)
• key durable state by conversation session_id while preserving raw task_id for sandbox/file resolution
• mark prior evidence stale after successful write_file / patch edits, using the edited file's project root when available
• match common real command spellings (pnpm test, bash scripts/run_tests.sh, uv run pytest) while avoiding echoed command text false positives
• bound storage growth: output summaries are capped, changed-path lists are capped, old per-session/workspace events are pruned, global pruning only deletes events no state row points at, and evidence naturally expires after 30 days

Notes

This is intentionally passive: it records evidence, not guarantees. A targeted test stays targeted, lint/typecheck/build are classified separately, and no stop-loop or desktop UI policy is added in this PR.

Test plan

• scripts/run_tests.sh tests/agent/test_verification_evidence.py tests/agent/test_coding_context.py -- -q

[github-actions]

🔎 Lint report: bb/verify-ledger vs origin/main

ruff

Total: 0 on HEAD, 0 on base (➖ 0)

🆕 New issues: none

✅ Fixed issues: none

Unchanged: 0 pre-existing issues carried over.

ty (type checker)

Total: 11261 on HEAD, 11259 on base (🆕 +2)

🆕 New issues (2):

Rule	Count
unresolved-attribute	2

First entries

tests/run_agent/test_credits_notices_toggle.py:76: [unresolved-attribute] unresolved-attribute: Unresolved attribute `_credits_session_start_micros` on type `AIAgent`
run_agent.py:2984: [unresolved-attribute] unresolved-attribute: Object of type `Self@get_credits_spent_micros` has no attribute `_credits_session_start_micros`

✅ Fixed issues (1):

Rule	Count
invalid-assignment	1

First entries

tests/run_agent/test_credits_notices_toggle.py:76: [invalid-assignment] invalid-assignment: Object of type `None` is not assignable to attribute `_credits_session_start_micros` of type `int`

Unchanged: 5940 pre-existing issues carried over.

Diagnostics are surfaced as warnings — this check never fails the build.

[tonydwb]

Code Review Summary

Verdict: Approved

Well-structured verification evidence ledger. The module is intentionally passive — recording evidence without making guarantees. Good design with bounded storage, scoped evidence (full vs targeted), and natural expiration. Comprehensive test coverage.

Changes:

• agent/verification_evidence.py: New 547-line module with SQLite-backed evidence ledger
• tests/agent/test_verification_evidence.py: 313 lines of tests
• tools/file_tools.py: 44 additions for file mutation tracking
• tools/terminal_tool.py: 29 additions for terminal result recording

LGTM.

Reviewed by Hermes Agent