mason-context 0.16.2 → 0.17.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -1,5 +1,30 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.17.0 — 2026-09-11
4
+
5
+ Mason can now retain an authorized advisory assessment through later sessions and commits, then reopen it when relevant evidence changes. Documentation audits cover nested instructions and README files with more precise reference and command scope.
6
+
7
+ - Record authorized advisory assessments through `review_advisory` and `mason audit review`. Preserve history, reuse decision reviews, retain closure through unrelated and metadata commits, and reopen when relevant evidence changes. Deferred or unavailable findings remain outstanding.
8
+ - Discover nested README and agent instruction files with their actual filename casing. Respect exclusions and preserve scope through local references, package commands, repair baselines, and automation caches.
9
+ - Keep never-tracked paths, ambiguous commands, and undocumented modules advisory. Recheckable candidates can resolve when their condition disappears; historical dependency and decision advisories still require review.
10
+ - Passive audits ignore directory aliases outside the source inventory while still refusing links used by selected source paths; fixes the audit abort reproduced on ripgrep's `HomebrewFormula` alias.
11
+ - Import individual native-validator outcomes through `mason-check-json`, including optional environment metadata. Preserve missing, skipped, failed, stale, and unknown evidence without running imported commands. Automatic compiler discovery and snippet execution remain future integrations.
12
+ - Update the pinned Markdown parser to `markdown-it` 14.3.1, covering its published denial-of-service fixes.
13
+ - Flush complete audit and repair JSON before exit when output is piped, including through the unified `mason audit` command.
14
+ - Clarify when agents should revise an existing lesson using its ID, including accepted records, instead of creating another proposal. Keep supported rationale and sources, distinguish genuinely separate lessons, and preserve review boundaries.
15
+ - Add knowledge-maintenance evaluations with a reduced R8 startup reproduction and a local publishing simulator. Fourteen live Codex sessions exercised reviewed revision, fresh-session reuse and controls; both Mason and ordinary notes passed. These development trials do not establish superiority over notes or production integration correctness. See the [results and limitations](bench/harness/knowledge/REVISION_GUIDANCE_RESULTS.md).
16
+
17
+ Run `mason upgrade` for standalone installations, or `npm install -g mason-context@0.17.0` for npm installations, then restart running assistants. Existing PATH-based integrations use the new binary on their next launch. In each project, rerun `mason setup --host codex` or `--host claude` to refresh the managed decision guidance and Git ignore rules for `.mason/reviews/`; review and commit those project changes. This preserves existing decisions and repair evidence and requires no decision-store migration. Setup does not approve findings or decision records.
18
+
19
+ ## 0.16.3 — 2026-09-09
20
+
21
+ - Coordinate overlapping hooks without holding the shared state lock during audit analysis. Reuse analysis for identical inputs, let changed inputs proceed independently, and revalidate before publishing results so older work cannot overwrite newer evidence.
22
+ - Retain independent execution receipts during parallel calls. Report repeated, durably recorded failures once per session until recovery; keep failures without receipts visible and preserve pre-edit evidence for shell and unknown tools.
23
+ - Accept decision bodies up to 2,500 characters, with a warning above the recommended 1,500. Preserve the full body, report precise validation limits, and shorten generated IDs at word boundaries where possible.
24
+ - Cover slow checks, overlapping CLI processes, failure recovery, and changing inputs with concurrency regressions. Simplify README badges for consistent rendering.
25
+
26
+ Run `mason upgrade` for standalone installations, then restart running assistants. Existing PATH-based project integrations do not need setup again for this release.
27
+
3
28
  ## 0.16.2 — 2026-09-09
4
29
 
5
30
  - Guard generated hooks when Mason is absent, and keep a checkout without local host setup quiet. Preserve active failure diagnostics and strict MCP startup; setup/status/teardown recognize the generated guard variants without project wrapper scripts.
package/README.md CHANGED
@@ -1,18 +1,22 @@
1
+ <div align="center">
2
+
1
3
  # Mason
2
4
 
3
- [![npm version](https://img.shields.io/npm/v/mason-context)](https://www.npmjs.com/package/mason-context)
4
- [![CI](https://img.shields.io/github/actions/workflow/status/adrianczuczka/mason/ci.yml?branch=main)](https://github.com/adrianczuczka/mason/actions/workflows/ci.yml)
5
- [![npm downloads](https://img.shields.io/npm/dm/mason-context)](https://www.npmjs.com/package/mason-context)
6
- [![license](https://img.shields.io/github/license/adrianczuczka/mason)](https://github.com/adrianczuczka/mason/blob/main/LICENSE)
7
- [![issues](https://img.shields.io/github/issues/adrianczuczka/mason)](https://github.com/adrianczuczka/mason/issues)
5
+ ### Your agents are creating tech debt. Mason helps you prevent it.
8
6
 
9
- ## Your AI agent is creating tech debt.
7
+ **Remember why decisions were made · Catch outdated guidance · Find what else needs updating**
10
8
 
11
- **Mason helps your coding agent catch what a patch leaves behind:** stale instructions, missed companion updates, and engineering decisions that need another look.
9
+ [![npm version](https://img.shields.io/npm/v/mason-context?style=flat-square&labelColor=334155&color=2563eb)](https://www.npmjs.com/package/mason-context)
10
+ [![CI](https://img.shields.io/github/actions/workflow/status/adrianczuczka/mason/ci.yml?branch=main&style=flat-square&labelColor=334155&label=CI)](https://github.com/adrianczuczka/mason/actions/workflows/ci.yml)
11
+ [![Monthly downloads](https://img.shields.io/npm/dm/mason-context?style=flat-square&labelColor=334155&color=2563eb)](https://www.npmjs.com/package/mason-context)
12
+ [![License](https://img.shields.io/github/license/adrianczuczka/mason?style=flat-square&labelColor=334155&color=64748b)](https://github.com/adrianczuczka/mason/blob/main/LICENSE)
12
13
 
13
- Works through MCP, with automatic documentation checks for **Codex** and **Claude Code**. No concept map required.
14
+ [![Works with Codex](https://img.shields.io/badge/Codex-supported-64748b?style=flat-square&labelColor=334155)](https://github.com/adrianczuczka/mason/blob/main/docs/setup.md)
15
+ [![Works with Claude Code](https://img.shields.io/badge/Claude%20Code-supported-64748b?style=flat-square&labelColor=334155)](https://github.com/adrianczuczka/mason/blob/main/docs/setup.md)
16
+ [![MCP compatible](https://img.shields.io/badge/MCP-compatible-64748b?style=flat-square&labelColor=334155)](https://github.com/adrianczuczka/mason/blob/main/docs/setup.md#other-clients)
17
+ [![Supported platforms](https://img.shields.io/badge/platforms-macOS%20%7C%20Linux%20%7C%20Windows-64748b?style=flat-square&labelColor=334155)](https://github.com/adrianczuczka/mason/blob/main/docs/distribution.md)
14
18
 
15
- Core checks run locally with no model calls or telemetry. Results become context for your existing assistant. [Data, network access, and team installation](docs/data-and-network.md) covers storage, optional networked features, and internal mirrors.
19
+ </div>
16
20
 
17
21
  ## Get started
18
22
 
@@ -54,7 +58,7 @@ mason audit --dir .
54
58
  mason review --dir . --base origin/main
55
59
  ```
56
60
 
57
- The audit checks claims in `AGENTS.md` and `CLAUDE.md`. The review checks committed changes against your chosen base. Both are read-only and need no model calls.
61
+ The audit checks claims in README and agent instruction files throughout the repository. The review checks committed changes against your chosen base. Both are read-only and need no model calls.
58
62
 
59
63
  ## What Mason catches
60
64