triss-coworker 0.43.0 → 0.44.1

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 (67) hide show
  1. package/ARCHITECTURE.md +5 -0
  2. package/CHANGELOG.md +128 -1
  3. package/README.md +71 -24
  4. package/SECURITY.md +23 -13
  5. package/bin/triss.js +11 -535
  6. package/docs/adr/2026-09-05-user-choice-and-easy-setup.md +98 -0
  7. package/docs/adr/README.md +5 -0
  8. package/docs/cli-reference.md +87 -3
  9. package/docs/configuration.md +81 -10
  10. package/docs/data-flows.md +8 -1
  11. package/docs/engines/crush.md +38 -4
  12. package/docs/engines/index.md +1 -1
  13. package/docs/engines/omp.md +29 -1
  14. package/docs/engines/opencode-go.md +1 -1
  15. package/docs/engines/opencode2.md +8 -0
  16. package/docs/generated/cli-reference.md +1214 -0
  17. package/docs/generated/mcp-reference.md +914 -0
  18. package/docs/getting-started.md +35 -14
  19. package/docs/integrations/jira.md +2 -1
  20. package/docs/integrations/linear.md +3 -2
  21. package/docs/mcp.md +40 -13
  22. package/docs/reliable-delegation-contract.md +16 -3
  23. package/docs/security-model.md +70 -15
  24. package/docs/troubleshooting.md +130 -6
  25. package/docs/usage-accounting.md +8 -3
  26. package/package.json +11 -5
  27. package/src/cli-program.js +564 -0
  28. package/src/coder-credential-proxy.js +341 -5
  29. package/src/coder-engine-registry.js +4 -1
  30. package/src/coder-engines/crush.js +96 -58
  31. package/src/coder-engines/omp.js +15 -7
  32. package/src/coder-providers.js +70 -95
  33. package/src/commands/ask.js +5 -0
  34. package/src/commands/chat.js +2 -1
  35. package/src/commands/coder.js +778 -285
  36. package/src/commands/commit-msg.js +2 -1
  37. package/src/commands/config.js +29 -429
  38. package/src/commands/init.js +26 -12
  39. package/src/commands/review.js +3 -1
  40. package/src/commands/status.js +48 -8
  41. package/src/commands/write.js +2 -1
  42. package/src/integrations/_contract.js +2 -1
  43. package/src/integrations/_registry.js +9 -2
  44. package/src/mcp/handlers.js +26 -7
  45. package/src/mcp/server.js +9 -1
  46. package/src/mcp/tools.js +26 -14
  47. package/src/migration/migrate.js +30 -0
  48. package/src/model-engine-adapters.js +26 -1
  49. package/src/model-projection-policy.js +37 -16
  50. package/src/model-runtime.js +18 -1
  51. package/src/model-selection.js +46 -3
  52. package/src/provider-config.js +6 -0
  53. package/src/provider-model-transport.js +290 -0
  54. package/src/provider-registry.js +6 -0
  55. package/src/secrets.js +205 -2
  56. package/src/setup/configuration.js +1052 -0
  57. package/src/setup/engines.js +637 -0
  58. package/src/setup/plan.js +589 -0
  59. package/src/setup/readiness.js +102 -0
  60. package/src/setup/wizard.js +1112 -0
  61. package/src/transport-registry.js +8 -1
  62. package/src/transports/openai-chat.js +2 -1
  63. package/src/transports/openai-responses.js +14 -1
  64. package/templates/claude-full.md +32 -19
  65. package/templates/claude.md +4 -4
  66. package/templates/codex-full.md +24 -13
  67. package/templates/codex.md +5 -5
package/ARCHITECTURE.md CHANGED
@@ -36,3 +36,8 @@ behavior change. Priority candidates are coder orchestration, model management,
36
36
  updater installation, MCP installation, and MCP handlers.
37
37
 
38
38
  Architecture decisions that change trust boundaries belong in `docs/adr/`.
39
+
40
+ The accepted owner decision on
41
+ [user choice, best-effort execution, and easy setup](docs/adr/2026-09-05-user-choice-and-easy-setup.md)
42
+ governs provider/engine availability and wizard design. Existing runtime gates
43
+ are not product requirements when they conflict with that decision.
package/CHANGELOG.md CHANGED
@@ -7,6 +7,133 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
7
7
 
8
8
  ## [Unreleased]
9
9
 
10
+ ## [0.44.1] — 2026-09-09
11
+
12
+ ### Fixed
13
+
14
+ - All nine open Dependabot security alerts (#21–#29) are closed in one
15
+ sweep. Root package: `hono` 4.13.7 (CVE-2026-84363 query-parser fragment
16
+ differential, CVE-2026-84364 `parseBody()` memory exhaustion,
17
+ CVE-2026-84365 `toSSG()` path traversal) and `js-yaml` 4.3.2
18
+ (CVE-2026-84375). Site: `astro` 7.3.2 (critical RCE via AVIF image
19
+ optimization, GHSA-26w7-cxv4-gfx2), `sharp` 0.35.4 (libheif
20
+ GHSA-g89c-p67h-r497, GHSA-2jg2-4ch7-h545), `js-yaml` 4.3.2, and `svgo`
21
+ 4.1.0 (CVE-2026-84369, CVE-2026-84370). `npm audit` reports 0
22
+ production vulnerabilities for the root package and the site.
23
+ - Carried dependency groups superseded from Dependabot: `@anthropic-ai/sdk`
24
+ ^0.124.0 (production) and the development-dependencies group (`keyv`
25
+ 5.6.0, `flat-cache` 6.1.23 lockfile-only).
26
+
27
+ ### Changed
28
+
29
+ - The 2026-09-08 documentation audit landed (D01–D16, A01–A03): generated
30
+ CLI and MCP references, security-model and getting-started updates, and
31
+ packaged-documentation acceptance in the publish pipeline.
32
+ - `triss-dsh-provider-bundle` is version-aligned at 0.44.1 and republished
33
+ unchanged; its OpenCode Zen, OpenCode Go, and Z.AI routes are unchanged.
34
+
35
+ ### Artifact integrity (0.44.1)
36
+
37
+ - `triss-dsh-provider-bundle-0.44.1.tgz` — sha256
38
+ `76b74f0092665c9f0a4e7cc1060f8f359ea9de04be59cd24557580d4f71ec3a1`
39
+ — `sha512-8/taMLUGjQjM2i4HME6gcZev+XycbhR6TmNv0zKtbPxT66MzVm98w3aZ1LPK3oNqqkPEMWz1yGMP5n1rGUdmNw==`
40
+ (computed with npm pack; the output is byte-deterministic).
41
+ - Root `triss-coworker-0.44.1.tgz` sha256 is reproducible via `npm pack` at
42
+ tag `v0.44.1`; the root tarball ships `CHANGELOG.md`, so its hash cannot
43
+ be recorded inside this file. Registry verification compares the packed
44
+ artifact with the published tarball byte-for-byte.
45
+
46
+ ## [0.44.0] — 2026-09-08
47
+
48
+ ### Added
49
+
50
+ - `triss config wizard` opens the Easy path directly: pick any canonical
51
+ provider, paste its key, choose which assistant host to connect (Claude
52
+ Code, Codex, both, or skip host wiring), review the summary before apply,
53
+ and finish with a first command to try. Reruns reuse the existing
54
+ configuration instead of resetting it.
55
+ - `triss config wizard --advanced` exposes the full tuning sections —
56
+ providers, execution (engine, effort, tri-state credential protection),
57
+ connections (MCP and agent rules selectable independently), integrations
58
+ (keep/unset with effective-source display), runtime (including pricing),
59
+ and in-place engine version pins.
60
+ - Targeted wizard setups configure one piece:
61
+ `triss config wizard <provider-id>`, `triss config wizard coder` (with
62
+ `--coder-provider`, `--coder-engine`, `--coder-protect-credentials`), or an
63
+ integration name.
64
+ - Headless apply: `triss config wizard --yes --agent none [--install]`
65
+ assembles a complete configuration from existing files, the environment,
66
+ and explicit flags; missing requirements exit non-zero before writing.
67
+ `triss init --setup` delegates to the same wizard before any rules write,
68
+ preserving init's target and scope intent.
69
+ - Every execution engine now serves every canonical provider: crush runs any
70
+ provider through a run-scoped config with `$ENV` credential references
71
+ (provider block seeded before `crush models use`), and opencode2, omp, and
72
+ crush execute non-coder projections (`ask`, `review`, `chat`, `write`,
73
+ `commit-msg`) best-effort with structured warnings naming the unverified
74
+ limitation. Provider and engine are independent choices for model-backed
75
+ commands, honored as selected with truthful warnings for best-effort
76
+ routes.
77
+ - `--protect-credentials` / `--no-protect-credentials` on model-backed
78
+ commands; MCP forwards `protect_credentials` as the same tri-state and
79
+ carries credential mode and disclosures in the run envelope.
80
+ - New persisted fields: `TRISS_CODER_PROVIDER` (coder-only provider default
81
+ that never rewrites `TRISS_DEFAULT_PROVIDER`), `TRISS_DEFAULT_EFFORT`,
82
+ `TRISS_CODER_EFFORT` (honored by every engine; crush discloses when a
83
+ provider declares no effort support), the `TRISS_PROTECT_CREDENTIALS` /
84
+ `TRISS_CODER_PROTECT_CREDENTIALS` tri-state (absent, true, or false — the
85
+ string `"false"` is never a truthy opt-in), and `TRISS_MODEL_TRANSPORTS`
86
+ (exact-model direct transport override map; a model without resolvable
87
+ direct metadata fails with the stable `TRISS_DIRECT_ENGINE_REQUIRED` code
88
+ and this remedy).
89
+
90
+ ### Changed
91
+
92
+ - Credential protection is a verbatim tri-state (absent/true/false) across
93
+ the CLI, MCP, and persisted configuration: `TRISS_PROTECT_CREDENTIALS` and
94
+ `TRISS_CODER_PROTECT_CREDENTIALS` persist explicit choices, and an explicit
95
+ false overrides a persisted true.
96
+ - The setup wizard no longer resets an existing provider/engine selection or
97
+ installs a Claude MCP server implicitly; Easy and Advanced share one setup
98
+ and resolution path, integrations gate readiness only when selected, and
99
+ readiness is re-derived from post-apply state.
100
+ - Engine installers run with a minimal environment allowlist (provider and
101
+ integration tokens never reach npm lifecycle scripts or downloaded
102
+ installers), curl|sh installers run as two argv spawns instead of a shell
103
+ string, and the exact install command is shown before Apply.
104
+ - Env-file writes are atomic (0600 temp file + rename after the CAS check);
105
+ host-plan status is tracked per concrete target; `config set` validates
106
+ provider and engine values; `triss status` reads `TRISS_CODER_ENGINE`
107
+ through the layered setup state.
108
+ - Migration gates the wizard on a `blocked` preflight state instead of
109
+ writing schema-2 over a conflict.
110
+
111
+ ### Fixed
112
+
113
+ - A debug path in the wizard's readiness diagnostics could print full
114
+ credentials to stderr; it is removed and privacy tests now guard the
115
+ route, the preserved value, and both former exposure points.
116
+ - MCP `protect_credentials: false` was indistinguishable from absent and
117
+ could not override a persisted true.
118
+ - OMP runs discarded usage/cost records; the coder transport protocol now
119
+ resolves from the same captured snapshot as the model, and child stderr
120
+ echoed in errors is masked.
121
+ - `triss init --setup` works in non-TTY shells (`--yes` registered).
122
+ - Root-endpoint provider URLs no longer gain a silent `/v1` in protected
123
+ runs; unverified cleanup no longer fails closed on a recycled pgid.
124
+
125
+
126
+ ### Artifact integrity (0.44.0)
127
+
128
+ - `triss-dsh-provider-bundle-0.44.0.tgz` — sha256
129
+ `c2e8ce70bb5964cdbcc16a42c574b6f85c142c36d905d066e042c0150f50b935`
130
+ — `sha512-l8rIL6i7vgukA7efR8Q2PvG8YBnAVRdnlWL338FgZmNt9r+/Efz8OgujnkFb39KLK1qe/sdHY+HntvcLZ9GrQw==`
131
+ (computed with npm pack; the output is byte-deterministic).
132
+ - Root `triss-coworker-0.44.0.tgz` sha256 is reproducible via `npm pack` at
133
+ tag `v0.44.0`; the root tarball ships `CHANGELOG.md`, so its hash cannot be
134
+ recorded inside this file. Registry verification compares the packed
135
+ artifact with the published tarball byte-for-byte.
136
+
10
137
  ## [0.43.0] — 2026-09-05
11
138
 
12
139
  ### Changed
@@ -39,7 +166,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
39
166
  effort, but help prose is not a compatibility gate. Capability errors now
40
167
  name missing options separately from below-minimum versions. Pre-suffixed
41
168
  public/configured OpenCode 2 models are rejected before engine side effects.
42
- - Added `TRISS_DEFAULT_ENGINE` so model-backed commands can persist a direct or verified OpenCode route. OpenCode projections install and verify an active primary agent pinned as `default_agent`; its deny-by-default context-only policy provides no ambient file, delegation, skill, or executable tools, and the final merged agent state is audited before credentials are forwarded. Engine warnings remain structured across MCP model tools, explicit credential protection is available, and unverified OpenCode 2, OMP, and Crush projections are rejected before launch.
169
+ - Added `TRISS_DEFAULT_ENGINE` so model-backed commands can persist a direct or verified OpenCode route. OpenCode projections install and verify an active primary agent pinned as `default_agent`; its deny-by-default context-only policy provides no ambient file, delegation, skill, or executable tools, and the final merged agent state is audited before credentials are forwarded. Engine warnings remain structured across MCP model tools, and explicit credential protection is available on every engine.
43
170
  - `triss_coder_run` retains the documented `protectCredentials` MCP spelling
44
171
  as a deprecated compatibility alias. It is OR-merged with
45
172
  `protect_credentials`, so either truthy spelling selects protected
package/README.md CHANGED
@@ -39,17 +39,43 @@ Delegation can reduce primary-agent context and inference costs, but savings dep
39
39
  npm install -g triss-coworker
40
40
  ```
41
41
 
42
- **Claude Code or Codex:** run `triss config wizard --standard`. The current Standard path configures the `openai-compatible` API key, main model, and small model, then asks whether to connect Claude Code, Codex, or both. It installs MCP and agent rules for your selection; this version has no Skip option in Standard. Use `triss config wizard --advanced` for another provider or granular setup.
42
+ `triss config wizard` opens the Easy path directly: pick a provider, paste its
43
+ key, choose the assistant hosts to register, and run your first command. It
44
+ reuses existing configuration instead of resetting it. `--advanced` exposes the
45
+ full sections (providers, execution, connections, integrations, runtime), and a
46
+ target argument (`triss config wizard opencode-go`, `triss config wizard coder`,
47
+ `triss config wizard jira`) sets up just that piece. Non-interactive shells add
48
+ `--yes` to apply a complete configuration assembled from existing files, the
49
+ environment, and explicit flags:
43
50
 
44
- **Terminal only:** configure the same profile without installing host integration. Omit values to enter them interactively; secret input is masked:
51
+ ```bash
52
+ triss config wizard --yes --agent none
53
+ ```
54
+
55
+ Connect Claude Code, Codex, or both:
56
+
57
+ ```bash
58
+ triss init --target claude --global --setup # runs the setup wizard first, then writes rules
59
+ triss init --target codex --global
60
+ triss mcp install --target claude --global
61
+ triss mcp install --target codex --global
62
+ ```
63
+
64
+ **Terminal only:** configure the same profile without installing host
65
+ integration. Omit values to enter them interactively; secret input is masked:
45
66
 
46
67
  ```bash
68
+ triss config set -g TRISS_OPENAI_COMPATIBLE_BASE_URL
47
69
  triss config set -g TRISS_OPENAI_COMPATIBLE_API_KEY
48
70
  triss config set -g TRISS_OPENAI_COMPATIBLE_MODEL
49
71
  triss config set -g TRISS_OPENAI_COMPATIBLE_SMALL_MODEL
50
72
  ```
51
73
 
52
- Check the credential and selected defaults with `triss status`. Existing provider and engine choices are not changed by these three profile-field commands; use `triss config wizard --advanced` to adjust them if needed.
74
+ `openai-compatible` is a configurable profile, not a provider detector: its
75
+ built-in endpoint is `https://api.deepseek.com/v1`. Set the base URL explicitly
76
+ when using another compatible endpoint. These four commands edit profile
77
+ fields; they do not reset an existing default provider or engine. Check the
78
+ effective configuration with `triss status` before the first request.
53
79
 
54
80
  For an installed host, run `triss mcp status` and restart that host's session. There is no need to repeat `triss init` or `triss mcp install` if the wizard already connected it. Terminal users skip this step.
55
81
 
@@ -102,19 +128,20 @@ Every provider has an endpoint, credential, `model` role, and `smallModel` role.
102
128
  4. global `~/.config/triss/.env`;
103
129
  5. registry defaults.
104
130
 
105
- `TRISS_DEFAULT_PROVIDER` selects the provider when a request omits one; `TRISS_DEFAULT_ENGINE` selects `direct`, `opencode`, `opencode2`, `omp`, or `crush` when it omits an engine. Direct CLI and MCP commands accept a native model id. Coder model options use `<provider>/<model-id>`. Read-only non-coder projection is currently verified only for `opencode`; `opencode2`, `omp`, and `crush` fail before launch instead of running a write-capable agent.
131
+ `TRISS_DEFAULT_PROVIDER` selects the provider when a request omits one; `TRISS_DEFAULT_ENGINE` selects `direct`, `opencode`, `opencode2`, `omp`, or `crush` when it omits an engine. The `--model` value is either a qualified `<canonical-provider>/<native-id>` selector or a bare native id; a bare id resolves against `--provider` when given, otherwise against the effective default provider. An explicit `--provider` that conflicts with a qualified model prefix is rejected, not silently replaced. Every engine can execute non-coder model projections (`ask`, `review`, `chat`, …); what differs per engine is the available protection, not permission to run. `opencode` and `direct` have a verified read-only projection; `opencode2` runs the same deny-everything projection agent through its run-scoped config surface, `omp` uses its run-private policy overlay, and `crush` runs single-agent with the restrict allowlist. Engines without a verified projection report their concrete limitation as a warning on the result — never as a refusal.
106
132
 
107
133
  Shared reasoning effort values:
108
134
 
109
135
  ```text
110
- minimal | low | medium | high | max
136
+ low | medium | high | xhigh | max
111
137
  ```
112
138
 
113
139
  Examples:
114
140
 
115
141
  ```bash
116
142
  triss ask --provider zai --model glm-5.2 --effort high \
117
- --paths src --question "Find correctness defects"
143
+ --paths 'src/**/*.js' \
144
+ --question "Find correctness defects. Cite file paths and line numbers."
118
145
 
119
146
  triss review --provider moonshot --model kimi-k3 --effort max
120
147
 
@@ -122,13 +149,21 @@ triss chat --provider openai-compatible --effort low \
122
149
  "Explain this error"
123
150
  ```
124
151
 
152
+ `--paths` accepts text files or quoted glob patterns. Directories are not read
153
+ recursively; an input like `--paths src` fails with a "no readable file
154
+ content" error before any model call. Quote globs so Triss, rather than the
155
+ shell, expands them.
156
+
125
157
  Omit `--model` to use the command's provider role. There are no public model presets.
126
158
 
127
159
  ## Configuration
128
160
 
129
161
  ```bash
130
- triss config wizard
131
- triss config wizard --local
162
+ triss config wizard # Easy setup (provider + key + hosts)
163
+ triss config wizard --advanced # full sections
164
+ triss config wizard opencode-go # targeted provider setup
165
+ triss config wizard coder --coder-engine omp --coder-provider moonshot
166
+ triss config wizard coder --coder-protect-credentials # persist TRISS_CODER_PROTECT_CREDENTIALS=true (proxy mode instead of default best-effort raw)
132
167
  triss config set TRISS_DEFAULT_PROVIDER zai
133
168
  triss config set TRISS_DEFAULT_ENGINE direct
134
169
  triss config get TRISS_ZAI_MODEL
@@ -152,16 +187,28 @@ triss ask --paths 'src/**/*.js' --question "Find correctness defects"
152
187
  triss review
153
188
  ```
154
189
 
155
- Explicit request flags still win. OpenCode-backed non-coder calls install and
156
- verify a run-scoped active primary `triss-readonly-projection` agent and pin
157
- `default_agent` to it before forwarding the selected credential. Its permission
158
- contract denies every tool by default because the complete request context is
159
- already supplied in the prompt; it never gains ambient file, shell, edit,
160
- skill, or delegation access. The process still runs as the current OS user and
161
- is not a filesystem sandbox. Add `--protect-credentials` to a model command
162
- when the selected credential can be kept behind the parent-owned proxy;
163
- raw-mode warnings are preserved in MCP structured results for every
164
- model-backed tool.
190
+ Explicit request flags still win. Engine-backed non-coder calls receive the
191
+ complete request context in the prompt. OpenCode installs and verifies a
192
+ run-scoped active primary `triss-readonly-projection` agent pinned as
193
+ `default_agent`; its permission contract denies every tool by default, so it
194
+ never gains ambient file, shell, edit, skill, or delegation access. The other
195
+ engines apply their best-effort equivalents (run-scoped config surface,
196
+ run-private policy overlay, or the restrict allowlist) and attach a warning
197
+ that names exactly what is not verified. The process still runs as the current
198
+ OS user and is not a filesystem sandbox. Best-effort tool-policy support does
199
+ not imply an automatic credential downgrade: a selected protected route fails
200
+ closed — before any credential-bearing spawn — when the raw key cannot be
201
+ contained by the checked credential boundary or when the proxy cannot start.
202
+ `--protect-credentials` requests the parent-owned credential proxy;
203
+ `--no-protect-credentials` is an explicit choice to run with the selected raw
204
+ credential and a disclosed limitation, and it overrides a persisted
205
+ `TRISS_PROTECT_CREDENTIALS=true` (or `TRISS_CODER_PROTECT_CREDENTIALS=true`)
206
+ choice for one run. For coder runs configured
207
+ through the wizard, `--coder-protect-credentials` / `--coder-no-protect-credentials`
208
+ persist `TRISS_CODER_PROTECT_CREDENTIALS=true` / `=false` (cannot be combined);
209
+ the coder-specific value takes precedence over the shared
210
+ `TRISS_PROTECT_CREDENTIALS`. Raw-mode and engine
211
+ warnings are preserved in MCP structured results for every model-backed tool.
165
212
 
166
213
  Provider fields:
167
214
 
@@ -206,7 +253,7 @@ Supported engines:
206
253
 
207
254
  - `opencode` — default OpenCode engine
208
255
  - `opencode2` — OpenCode 2 beta, current-or-newer compatibility
209
- - `crush` — Z.A.I-only engine
256
+ - `crush` — provider-neutral single-envelope engine (any canonical provider)
210
257
  - `omp` — native Oh My Pi adapter
211
258
 
212
259
  OpenCode 2 has a supported floor of `0.0.0-beta-19059` and accepts every newer
@@ -219,8 +266,8 @@ Setup examples:
219
266
  ```bash
220
267
  triss coder init --engine opencode --provider openai-compatible
221
268
  triss coder init --engine opencode2 --provider opencode-zen
222
- triss coder init --engine crush --provider zai
223
- triss coder init --engine omp --provider opencode-go
269
+ triss coder init --engine crush --provider opencode-go
270
+ triss coder init --engine omp --provider moonshot
224
271
  ```
225
272
 
226
273
  Run examples:
@@ -237,7 +284,7 @@ triss coder run --isolate --session auth-fix \
237
284
  "Fix the authentication bug and run focused checks"
238
285
  ```
239
286
 
240
- Each run forwards only the selected provider credential. Protected routes use parent-owned loopback credential mediation and fail closed when an engine projection or endpoint cannot be audited. `--isolate` uses `.triss/wt/<slug>` for a reviewable worktree.
287
+ Each run forwards only the selected provider credential. Protected routes use parent-owned loopback credential mediation and fail closed when a protected route cannot actually contain the real key; an explicit `--no-protect-credentials` runs crush raw through the same run-scoped config with the selected credential (best-effort, warned). `--isolate` uses `.triss/wt/<slug>` for a reviewable worktree.
241
288
 
242
289
  The `opencode` V1 engine preserves native OpenCode routing for Zen and Go. In
243
290
  protected mode, only `User-Agent` plus session, request, and client identity
@@ -255,7 +302,7 @@ Engine details:
255
302
 
256
303
  If the wizard has not already connected your host, choose Claude Code or Codex in the [host-connection guide](https://triss.work/docs/getting-started/#step-4). Only run setup for the host you intend to use.
257
304
 
258
- Core tools include `triss_ask`, `triss_chat`, `triss_fetch`, `triss_review`, `triss_write`, `triss_commit_msg`, `triss_status`, and the migration/update surfaces. Coder tools appear when any canonical provider credential is configured. Tracker tools appear only when their integration credential is ready.
305
+ Core tools include `triss_ask`, `triss_chat`, `triss_fetch`, `triss_review`, `triss_review_shard`, `triss_write`, `triss_commit_msg`, and `triss_status`. Updates and migration stay CLI-side: the MCP server emits passive update notices but exposes no update or migration tool. Coder tools appear when any canonical provider credential is configured. Tracker tools appear only when their integration credential is ready.
259
306
 
260
307
  The MCP schemas use the same `provider`, `model`, `effort`, and engine contracts as the CLI. Full reference: [MCP](https://github.com/ayleen/triss-coworker/blob/main/docs/mcp.md).
261
308
 
@@ -284,7 +331,7 @@ Usage records preserve provider, model, token-class provenance, billing mode, an
284
331
  ## Updates
285
332
 
286
333
  ```bash
287
- triss update --check
334
+ triss update
288
335
  triss update --apply
289
336
  triss update --rollback
290
337
  ```
package/SECURITY.md CHANGED
@@ -1,8 +1,8 @@
1
1
  # Security
2
2
 
3
3
  Triss is a local CLI and MCP server that can read files, fetch URLs, call
4
- tracker APIs, and send selected context to a DeepSeek-compatible model. That
5
- makes the trust boundary worth spelling out plainly.
4
+ tracker APIs, and send selected context to the model provider you configure.
5
+ That makes the trust boundary worth spelling out plainly.
6
6
 
7
7
  ## Reporting vulnerabilities
8
8
 
@@ -22,8 +22,8 @@ advisory unless they ask to remain anonymous.
22
22
 
23
23
  ## What leaves your machine
24
24
 
25
- Model-backed commands send the requested prompt and selected corpus to the
26
- configured OpenAI-compatible endpoint:
25
+ Model-backed commands send the requested prompt and selected context to the
26
+ resolved endpoint of your selected canonical provider:
27
27
 
28
28
  - `triss ask`
29
29
  - `triss chat`
@@ -84,21 +84,31 @@ content is never written to this log** — metadata only. If working-directory
84
84
  paths are themselves sensitive (client names in folder names), set
85
85
  `TRISS_USAGE_LOG_CWD=0`, or disable the log entirely with
86
86
  `TRISS_USAGE_LOG=0`. The file rotates once past `TRISS_USAGE_LOG_MAX_BYTES`;
87
- the generated defaults table in [docs/configuration.md](docs/configuration.md#tunables)
87
+ the generated defaults table in
88
+ [docs/configuration.md](docs/configuration.md#network-and-usage-controls)
88
89
  is the source of truth. Delete it at any time with `triss usage --reset`.
89
90
 
90
91
  ## Data residency and GDPR
91
92
 
92
93
  Triss is a local tool, not a hosted service — it stores none of your data
93
94
  server-side and has no subprocessors of its own. The party that processes
94
- your prompts is **whatever model endpoint you configure**. The default is
95
- DeepSeek (`api.deepseek.com`); if your compliance posture requires an EU- or
96
- US-resident processor, a signed DPA, or a zero-retention guarantee, point
97
- `TRISS_WORKER_BASE_URL` at a provider that offers one (Azure OpenAI,
98
- AWS Bedrock, Mistral, or a self-hosted vLLM/Ollama endpoint — see the
99
- provider recipes in the README) and set the model names accordingly. Your
100
- organisation's agreement with that provider is the controlling document;
101
- Triss adds no additional data flows on top of it.
95
+ your prompts is **whatever model endpoint you configure**.
96
+
97
+ Model-backed commands send their selected context to the resolved provider
98
+ endpoint. For the `openai-compatible` profile, configure
99
+ `TRISS_OPENAI_COMPATIBLE_BASE_URL`, `TRISS_OPENAI_COMPATIBLE_API_KEY`, and the
100
+ profile's model fields. Other canonical providers use their own endpoint
101
+ fields (see the provider field table in
102
+ [docs/configuration.md](docs/configuration.md#canonical-providers)).
103
+ `TRISS_WORKER_BASE_URL` is a pre-0.42 migration input, not a current setup
104
+ instruction; the migration reference in
105
+ [src/migration/legacy-inventory.js](src/migration/legacy-inventory.js) maps it
106
+ to the canonical field. Triss does not enforce a provider's retention,
107
+ training, residency, or contractual policies — your organisation's agreement
108
+ with that provider is the controlling document. Triss does not add undocumented
109
+ data flows on top of the configured provider; see the complete flow table in
110
+ [docs/data-flows.md](docs/data-flows.md), including integrations, approved
111
+ setup downloads, and the credential-free passive update check.
102
112
 
103
113
  ## Supply chain
104
114