contextos-agents 2.3.0 → 2.3.1

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.
Files changed (3) hide show
  1. package/README.md +31 -21
  2. package/bin/index.js +1 -1
  3. package/package.json +89 -89
package/README.md CHANGED
@@ -59,17 +59,19 @@ npx contextos-agents init --all # Install all 36 catalog skills at
59
59
 
60
60
  ### Options
61
61
 
62
- Try the [small local demo](examples/quickstart/README.md) to install the current
63
- candidate in a new folder, select a TypeScript skill, and add a team rule without
62
+ Try the [small local demo](examples/quickstart/README.md) to install the package in a new folder, select a TypeScript skill, and add a team rule without
64
63
  calling a model API. The [five-minute guide](docs/product/onboarding.md) explains
65
64
  the same workflow for an existing project.
66
65
 
67
- For the current R2 candidates, see the [release preparation status](docs/R2_RELEASE_PREPARATION.md)
66
+ Core 2.3.1 / MCP 0.4.1 are being prepared as a maintenance release; they are not yet published.
67
+ See the [candidate checklist](docs/PATCH_RELEASE_2.3.1.md) and
68
+ [live client check](docs/LIVE_CLIENT_CHECK_RU.md).
69
+ Core 2.3.0 and MCP 0.4.0 remain the published pair. See the [release status](docs/R2_RELEASE_PREPARATION.md)
68
70
  and [upgrade/checkpoint rollback](docs/COMPACT_CONTEXT_MIGRATION.md). The
69
- [release manifest](docs/evidence/release-2.3.json) records the verified source,
70
- cross-platform CI and candidate archive identities. The client pilot and
71
- publication decision remain pending. Internal plans, local API probes
72
- and raw logs are excluded from the public release surface.
71
+ [release manifest](docs/evidence/release-2.3.json) records the released source,
72
+ cross-platform CI and archive identities. Automatic client routing and the external
73
+ pilot remain unverified. Internal plans, local API probes and raw logs are excluded
74
+ from the public release surface.
73
75
 
74
76
  ```bash
75
77
  npx contextos-agents --help # Show all options
@@ -88,13 +90,13 @@ npx contextos-agents --skip-compile # Skip auto-compilation step
88
90
 
89
91
  Modern development teams face fragmented AI tooling: engineers use Cursor, Claude Code, GitHub Copilot, Gemini, Zed, and Aider. Each tool requires its own proprietary rules format, leading to configuration drift, contradictory standards, and unvetted AI slop (`// TODO`, leaked secrets).
90
92
 
91
- Artificially truncating skills to save tokens degrades model reasoning and induces hallucinations. Instead, ContextOS ensures that agents receive complete, high-fidelity engineering context from a single version-controlled source.
93
+ ContextOS preserves selected skill bodies and reports budget overflow instead of silently truncating required instructions. Whether a client loads and follows those rules requires separate verification. The current [calibration](docs/BENCHMARK_RESULTS.md) found no quality advantage over vanilla on its test corpus.
92
94
 
93
95
  **ContextOS is not another coding agent.** It is the deterministic context compiler and policy engine for the agents your team already uses.
94
96
 
95
97
  ### The Three Pillars
96
98
 
97
- 1. **Portable (Multi-Agent):** Define your engineering skills once in standard Markdown. ContextOS compiles native configurations for all supported agents (Gemini, Claude Code, Cursor, Copilot, Aider, and Zed).
99
+ 1. **Portable (Multi-Agent):** Define your engineering skills once in standard Markdown. ContextOS generates agent-specific exports (Gemini, Claude Code, Cursor, Copilot, Aider, and Zed); native loading and manual templates are distinguished in the compatibility matrix.
98
100
  2. **High-Fidelity & Focused:** The resolver maps domain skills to relevant tasks without lossy truncation, delivering rich, complete context to the model.
99
101
  3. **Verifiable in CI:** Lockfile v2 provenance, dual-hash verification, and CI quality gates detect configuration drift and enforce quality guardrails before merge.
100
102
 
@@ -111,15 +113,16 @@ contextos resolve "review authentication changes" \
111
113
  --explain
112
114
  ```
113
115
 
114
- Selected:
115
- security explicit task match
116
- engineering-workflow required dependency
116
+ Example excerpt from the repository configuration (selection and scores depend on installed skills, files and profile):
117
117
 
118
- Excluded:
119
- context-manager domain relevance filter
118
+ ```text
119
+ Selected Skills:
120
+ ✓ engineering-workflow score: 100 tokens: ~418 (foundation: Core skill)
121
+ ✓ security score: 185 tokens: ~2237 (safety_required: Touched file "src/auth/session.ts" matches glob "**/*auth*" (+25); safety_required: Touched file "src/auth/session.ts" matches pattern (+60); safety_required: Required safety guidance for high risk task)
122
+ ✓ typescript score: 60 tokens: ~1000 (file_glob: Touched file "src/auth/session.ts" matches pattern (+60))
123
+ ```
120
124
 
121
- Risk: high
122
- Context status: complete and verified
125
+ Selection does not imply installation or activation. If TypeScript is not installed, the CLI also reports `CTX_SELECTED_SKILL_UNAVAILABLE`; install it with `contextos skill add typescript` before runtime assembly.
123
126
 
124
127
  ## Dynamic Skill Resolution & Unified CLI (`contextos` / `ctx.js`)
125
128
 
@@ -132,8 +135,9 @@ ContextOS provides a unified CLI (`contextos` or `npx contextos-agents`) and loc
132
135
  contextos resolve "Build an accessible modal component with React and Tailwind"
133
136
 
134
137
  # Output:
135
- # [DOMAIN: Frontend] [PHASE: Build] [ROLE: Senior Developer]
136
- # Skills loaded: ponytail-mindset, engineering-workflow, gemini-precision
138
+ # [DOMAIN: Frontend] [PHASE: Build] [ROLE: Senior Developer] [MODE: CHANGE] [LENSES: accessibility] [RISK: standard]
139
+ # Skills loaded: ponytail-mindset, engineering-workflow, react, web-accessibility, ui-ux-pro
140
+ # Selection declaration only; skill bodies are read by the consuming client or prompt assembler.
137
141
 
138
142
  # Resolve with full evidence scoring explanation:
139
143
  contextos resolve "security review" --files apps/web/app/login/page.tsx --explain
@@ -227,13 +231,17 @@ jobs:
227
231
  runs-on: ubuntu-latest
228
232
  steps:
229
233
  - uses: actions/checkout@v4
230
- - uses: kok-o/contextos-agents/.github/actions/contextos-gate@v2.2.0
234
+ - uses: kok-o/contextos-agents/.github/actions/contextos-gate@v2.3.1
231
235
  with:
232
- version: '2.2.0' # Pinned version of contextos-agents runner
236
+ version: '2.3.1' # Pinned version of contextos-agents runner
233
237
  adapters: 'all' # Adapters to verify (or specific: 'cursor', 'claude')
234
238
  working-directory: '.' # Project root directory
235
239
  ```
236
240
 
241
+ The example targets the forthcoming v2.3.1 release; use it after that tag and package are published.
242
+ Until then, keep both pins at v2.3.0 / 2.3.0. The existing `v2.3.0` action tag defaults to CLI 2.2.0,
243
+ so retain the explicit `version`. Updating the action source does not change an existing tag.
244
+
237
245
  The action executes the verified ContextOS quality gate in-process from the pinned package version, verifying generated AI adapter configs against source skills without executing untrusted scripts from pull requests, and without requiring a Node.js project or running `npm test`.
238
246
 
239
247
  ## Optional MCP integration (Beta)
@@ -243,10 +251,12 @@ The MCP server is a separate beta package. It is not part of the stable `context
243
251
  Install it separately if you want to try the beta integration:
244
252
 
245
253
  ```bash
246
- npm install --save-dev @contextos/mcp
254
+ npm install --save-dev contextos-mcp
247
255
  npx contextos-mcp --dir .
248
256
  ```
249
257
 
258
+ The former `@contextos/mcp` 0.3.x package is historical; use `contextos-mcp` for current releases.
259
+
250
260
  The MCP server is read-only by default. Runtime execution remains experimental and is outside the stable core scope.
251
261
 
252
262
  ## Security - Third-Party Skills
package/bin/index.js CHANGED
@@ -100,7 +100,7 @@ Options:
100
100
  --project <path> Specify target project root directory (default: current directory)
101
101
  --target <adapters> Target adapter(s) to verify or export (default: all)
102
102
  --github-annotations Emit GitHub Actions workflow commands and step summary
103
- --with-mcp, --mcp [DEPRECATED] Use the separate @contextos/mcp package instead
103
+ --with-mcp, --mcp [DEPRECATED] Use the separate contextos-mcp package instead
104
104
  --skip-compile Skip running ctx.js export after installation
105
105
  --add-skill <ref> Install a community plugin skill after setup
106
106
 
package/package.json CHANGED
@@ -1,91 +1,91 @@
1
1
  {
2
- "name": "contextos-agents",
3
- "version": "2.3.0",
4
- "description": "Deterministic context and policy compiler for supported AI coding agents.",
5
- "bin": {
6
- "contextos": "./bin/index.js",
7
- "contextos-agents": "./bin/index.js",
8
- "koko-contextos-agents": "./bin/index.js"
9
- },
10
- "files": [
11
- "NOTICE",
12
- "bin",
13
- "catalog",
14
- "registry.json",
15
- "registry.schema.json",
16
- "registry.v2.schema.json",
17
- ".agents/AGENTS.md",
18
- ".agents/skills.json",
19
- ".agents/ctx.js",
20
- ".agents/validate.js",
21
- ".agents/plugins.js",
22
- ".agents/profiles.js",
23
- ".agents/resolver.js",
24
- ".agents/resolver",
25
- ".agents/rules",
26
- ".agents/doctor.js",
27
- ".agents/stats.js",
28
- ".agents/watch.js",
29
- ".agents/skills-index.json",
30
- ".agents/adapters",
31
- ".agents/core",
32
- ".agents/generated",
33
- ".agents/plugins/contextos",
34
- ".agents/compiled",
35
- ".agents/compiler",
36
- ".agents/schemas",
37
- ".agents/workspace",
38
- ".agents/filesystem",
39
- ".agents/transaction-core",
40
- ".agents/customization-dx.js"
41
- ],
42
- "scripts": {
43
- "build": "node .agents/ctx.js compile \u0026\u0026 node .agents/ctx.js export all",
44
- "compile": "node .agents/ctx.js compile",
45
- "export": "node .agents/ctx.js export gemini",
46
- "validate": "node .agents/ctx.js validate",
47
- "validate:catalog": "node .agents/ctx.js validate --catalog",
48
- "lint:md": "markdownlint README.md CONTRIBUTING.md GUIDE.md \".agents/core/skills/**/*.md\" --ignore \".agents/generated/**\" --config .markdownlint.json",
49
- "check:secrets": "node scripts/check-secrets.js --all",
50
- "check:consumer": "node scripts/check-release-install.cjs",
51
- "check:migration": "node scripts/check-release-migration.cjs",
52
- "check:release-surface": "node scripts/check-release-surface.cjs",
53
- "setup:hooks": "node scripts/install-hooks.js",
54
- "watch": "node .agents/watch.js",
55
- "test:skills": "node scripts/verify-skill-examples.js",
56
- "test": "node --test --test-concurrency=1 tests/install.test.js tests/lockfile.test.js tests/safe-writer.test.js tests/adapter-safety.test.js tests/update.test.js tests/uninstall.test.js tests/plugin-security.test.js tests/export.test.js tests/skills.test.js tests/validate.test.js tests/plugins.test.js tests/benchmark.test.js tests/benchmark-api.test.js tests/profile.test.js tests/resolver.test.js tests/cli-registry.test.js tests/profile-enforcement.test.js tests/doctor-hardening.test.js tests/manifest-compiler.test.js tests/canonical-resolver.test.js tests/workspace-graph.test.js tests/profiles-v2.test.js tests/safe-path.test.js tests/project-lock.test.js tests/journaled-transaction.test.js tests/pure-adapters.test.js tests/prompt-skill-system.test.js tests/runtime-state-machine.test.js tests/verification-reviewer-pipeline.test.js tests/durable-concurrency-locks.test.js tests/transactional-git.test.js tests/sandbox-execution.test.js tests/plugin-supply-chain.test.js tests/platform-hardening.test.js tests/customization-dx.test.js tests/claims-governance.test.js tests/benchmark-v2.test.js tests/init-engine.test.js tests/mutation-paths.test.js tests/tarball-smoke.test.js tests/catalog-export-integration.test.js tests/consumer-gate.test.js tests/adapter-compatibility.test.js tests/consumer-init.test.js tests/skill-examples.test.js tests/scan.test.js tests/hooks.test.js tests/recovery-cli.test.js tests/doctor-v2.test.js tests/resolver-parity.test.js tests/watch-coalescing.test.js tests/consumer-stabilization.test.js tests/focused-context.test.js tests/spend-guard.test.js tests/maintenance-benchmark.test.js tests/luna-agent-benchmark.test.js tests/release-surface.test.js",
57
- "prepublishOnly": "npm run validate \u0026\u0026 npm run build"
58
- },
59
- "keywords": [
60
- "context-compiler",
61
- "context-governance",
62
- "policy-as-code",
63
- "ai-agents",
64
- "coding-agents",
65
- "developer-tools",
66
- "cursor",
67
- "claude-code",
68
- "github-copilot",
69
- "gemini",
70
- "zed",
71
- "contextos"
72
- ],
73
- "author": "koko-o",
74
- "license": "Apache-2.0",
75
- "repository": {
76
- "type": "git",
77
- "url": "https://github.com/kok-o/contextos-agents.git"
78
- },
79
- "homepage": "https://github.com/kok-o/contextos-agents#readme",
80
- "bugs": {
81
- "url": "https://github.com/kok-o/contextos-agents/issues"
82
- },
83
- "engines": {
84
- "node": "\u003e=22.0.0"
85
- },
86
- "devDependencies": {
87
- "esbuild": "^0.28.2",
88
- "markdownlint-cli": "^0.43.0",
89
- "yaml": "^2.9.0"
90
- }
2
+ "name": "contextos-agents",
3
+ "version": "2.3.1",
4
+ "description": "Deterministic context and policy compiler for supported AI coding agents.",
5
+ "bin": {
6
+ "contextos": "./bin/index.js",
7
+ "contextos-agents": "./bin/index.js",
8
+ "koko-contextos-agents": "./bin/index.js"
9
+ },
10
+ "files": [
11
+ "NOTICE",
12
+ "bin",
13
+ "catalog",
14
+ "registry.json",
15
+ "registry.schema.json",
16
+ "registry.v2.schema.json",
17
+ ".agents/AGENTS.md",
18
+ ".agents/skills.json",
19
+ ".agents/ctx.js",
20
+ ".agents/validate.js",
21
+ ".agents/plugins.js",
22
+ ".agents/profiles.js",
23
+ ".agents/resolver.js",
24
+ ".agents/resolver",
25
+ ".agents/rules",
26
+ ".agents/doctor.js",
27
+ ".agents/stats.js",
28
+ ".agents/watch.js",
29
+ ".agents/skills-index.json",
30
+ ".agents/adapters",
31
+ ".agents/core",
32
+ ".agents/generated",
33
+ ".agents/plugins/contextos",
34
+ ".agents/compiled",
35
+ ".agents/compiler",
36
+ ".agents/schemas",
37
+ ".agents/workspace",
38
+ ".agents/filesystem",
39
+ ".agents/transaction-core",
40
+ ".agents/customization-dx.js"
41
+ ],
42
+ "scripts": {
43
+ "build": "node .agents/ctx.js compile && node .agents/ctx.js export all",
44
+ "compile": "node .agents/ctx.js compile",
45
+ "export": "node .agents/ctx.js export gemini",
46
+ "validate": "node .agents/ctx.js validate",
47
+ "validate:catalog": "node .agents/ctx.js validate --catalog",
48
+ "lint:md": "markdownlint README.md CONTRIBUTING.md GUIDE.md \".agents/core/skills/**/*.md\" --ignore \".agents/generated/**\" --config .markdownlint.json",
49
+ "check:secrets": "node scripts/check-secrets.js --all",
50
+ "check:consumer": "node scripts/check-release-install.cjs",
51
+ "check:migration": "node scripts/check-release-migration.cjs",
52
+ "check:release-surface": "node scripts/check-release-surface.cjs",
53
+ "setup:hooks": "node scripts/install-hooks.js",
54
+ "watch": "node .agents/watch.js",
55
+ "test:skills": "node scripts/verify-skill-examples.js",
56
+ "test": "node --test --test-concurrency=1 tests/install.test.js tests/lockfile.test.js tests/safe-writer.test.js tests/adapter-safety.test.js tests/update.test.js tests/uninstall.test.js tests/plugin-security.test.js tests/export.test.js tests/skills.test.js tests/validate.test.js tests/plugins.test.js tests/benchmark.test.js tests/benchmark-api.test.js tests/profile.test.js tests/resolver.test.js tests/cli-registry.test.js tests/profile-enforcement.test.js tests/doctor-hardening.test.js tests/manifest-compiler.test.js tests/canonical-resolver.test.js tests/workspace-graph.test.js tests/profiles-v2.test.js tests/safe-path.test.js tests/project-lock.test.js tests/journaled-transaction.test.js tests/pure-adapters.test.js tests/prompt-skill-system.test.js tests/runtime-state-machine.test.js tests/verification-reviewer-pipeline.test.js tests/durable-concurrency-locks.test.js tests/transactional-git.test.js tests/sandbox-execution.test.js tests/plugin-supply-chain.test.js tests/platform-hardening.test.js tests/customization-dx.test.js tests/claims-governance.test.js tests/benchmark-v2.test.js tests/init-engine.test.js tests/mutation-paths.test.js tests/tarball-smoke.test.js tests/catalog-export-integration.test.js tests/consumer-gate.test.js tests/adapter-compatibility.test.js tests/consumer-init.test.js tests/skill-examples.test.js tests/scan.test.js tests/hooks.test.js tests/recovery-cli.test.js tests/doctor-v2.test.js tests/resolver-parity.test.js tests/watch-coalescing.test.js tests/consumer-stabilization.test.js tests/focused-context.test.js tests/spend-guard.test.js tests/maintenance-benchmark.test.js tests/luna-agent-benchmark.test.js tests/release-surface.test.js",
57
+ "prepublishOnly": "npm run validate && npm run build"
58
+ },
59
+ "keywords": [
60
+ "context-compiler",
61
+ "context-governance",
62
+ "policy-as-code",
63
+ "ai-agents",
64
+ "coding-agents",
65
+ "developer-tools",
66
+ "cursor",
67
+ "claude-code",
68
+ "github-copilot",
69
+ "gemini",
70
+ "zed",
71
+ "contextos"
72
+ ],
73
+ "author": "koko-o",
74
+ "license": "Apache-2.0",
75
+ "repository": {
76
+ "type": "git",
77
+ "url": "https://github.com/kok-o/contextos-agents.git"
78
+ },
79
+ "homepage": "https://github.com/kok-o/contextos-agents#readme",
80
+ "bugs": {
81
+ "url": "https://github.com/kok-o/contextos-agents/issues"
82
+ },
83
+ "engines": {
84
+ "node": ">=22.0.0"
85
+ },
86
+ "devDependencies": {
87
+ "esbuild": "^0.28.2",
88
+ "markdownlint-cli": "^0.43.0",
89
+ "yaml": "^2.9.0"
90
+ }
91
91
  }