Prepare v0.9.1 release metadata (#81)

Co-authored-by: admin <admin@SH-TuShaokun.local>

Bump version to 0.8.0, update README with v0.8 features (#76)

New in v0.8.0:
- Chain mode (--chain) for sequential multi-agent analysis
- Per-provider perspectives (--perspectives-json)
- Consensus badges ([N/M agree] / [confirmed by N/M])
- Session retry with error classification and exponential backoff
- Session result retrieval (session result <name> <id>)
- Critical nowait data loss fix

Co-authored-by: admin <admin@SH-TuShaokun.local>

Bump version to 0.7.0, update README with v0.7 features (#74)

Co-authored-by: admin <admin@SH-TuShaokun.local>

Bump version to 0.6.0 (#71)

Co-authored-by: admin <admin@SH-TuShaokun.local>

Bump version to 0.5.0 (#65)

Co-authored-by: admin <admin@SH-TuShaokun.local>

feat: cross-session memory bridge (Phase 1-4) (#60)

* feat: add single-slot run hooks for MCO lifecycle events

* feat: add --memory/--space CLI flags and hook wiring in run_review

* feat: add bridge skeleton with finding_hash and PersistedFinding type

* feat: add evermemos MCP client wrapper with optional dependency

* feat: implement bridge core with pre_run/post_run and finding merge semantics

* test: add integration tests for hook wiring and real bridge path

* test: verify memory misconfiguration error messages are actionable

* fix: use uvx evermemos-mcp (PyPI) instead of npx npm placeholder

* fix: reject colon in --space (slug only), add uvx-missing test

* feat: add passive confirmation logic for finding lifecycle

* feat: add forget cleaner for rejected findings

* feat: add task type auto-classification from prompt and findings

* feat: add tech stack detection from project file markers

* feat: add request_status polling for write confirmation

* feat: add agent reliability scoring with cross-validation tracking

* feat: add cold-start weight mixing for agent reliability

* feat: add finding confidence calculation with three-factor weighting

* feat: wire passive confirm and forget into post_run hook

* feat: integrate confidence, classification, and stack detection into bridge hooks

Wire Phase 2-4 modules into the bridge pre_run and post_run hooks:
- Stack detection and agent weight retrieval in pre_run
- Confidence calculation before finding persistence in post_run
- Task classification and agent score persistence in post_run
- Status polling for critical/high severity findings in post_run
- Confidence grade labels in injected prompt for known-open findings

* feat: add tech stack aggregate scoring for cold-start priors

Add update_stack_aggregate() to scoring.py that reads existing scores
from a stack-level evermemos space, merges with new per-repo scores,
and writes back. Wire into core.py _post_run_impl() after repo-specific
agent scores. This enables cold-start priors for new repos sharing the
same tech stack.

* feat: add mco findings subcommand for listing and confirming findings

* feat: add mco memory subcommand for agent-stats, priors, and status

* test: add end-to-end integration test for full Phase 1-4 pipeline

* fix: resolve 5 correctness issues from code review

1. Canonical latest view: deduplicate findings/scores by key across
   all read paths (core, findings_cli, memory_cli)
2. Accurate run_count: count [MCO-RUN-MARKER] entries in context space
   instead of finding entry count
3. Category-aware _load_agent_rates: filter by task_category or average
   across categories when no filter specified
4. memory_id injection: extract id from fetch_history outer item, not
   from serialized content JSON
5. Per-finding changed_files: each finding checks files changed since
   its own last_seen_commit, not a union across all commits

Add 17 targeted tests covering all 5 fixes.

* fix: wire category parameter through show_priors to _load_agent_rates

show_priors() accepted a category argument but never forwarded it to
_load_agent_rates(), making the category filter decorative. Now all
three _load_agent_rates() calls (repo, stack, global) receive the
category so users get category-filtered priors as intended.

* docs: add cross-session memory section to README, bump to v0.4.0

Add opt-in --memory feature documentation with usage examples,
subcommands, and architecture overview. Keep it concise at the
bottom of the README. Bump version to 0.4.0 for the memory bridge
feature release.

* fix: replace pytest.approx with unittest assertions for CI compatibility

CI runs python3 -m unittest discover without pytest installed.
Convert test_cold_start.py and test_confidence.py to use
unittest.TestCase with assertAlmostEqual/assertEqual.

---------

Co-authored-by: admin <admin@SH-TuShaokun.local>

Bump version to 0.3.5 (0.3.4 was already published on npm) (#58)

Co-authored-by: admin <admin@SH-TuShaokun.local>

chore(release): v0.3.3 (#44)

Bump npm/PyPI package versions to 0.3.3 and cut changelog
entries for doctor, markdown-pr/sarif formats, token usage, findings
dedupe, and synthesis output support.

Merge pull request #34 from mco-org/chore/release-v0.3.1

chore(release): cut v0.3.1