kolega-code 0.13.0__tar.gz → 0.14.0__tar.gz

{kolega_code-0.13.0 → kolega_code-0.14.0}/.gitignore

@@ -14,3 +14,4 @@ build/
 .coverage
 coverage.xml
 htmlcov/
+docs/kolega_reskin_project/

{kolega_code-0.13.0 → kolega_code-0.14.0}/.pre-commit-config.yaml

@@ -15,3 +15,13 @@ repos:
       - id: check-yaml
       - id: debug-statements
       - id: detect-private-key
+
+  - repo: local
+    hooks:
+      - id: pyright
+        name: pyright type checker
+        entry: uv run pyright
+        language: system
+        types: [python]
+        pass_filenames: false
+        require_serial: true

{kolega_code-0.13.0 → kolega_code-0.14.0}/CHANGELOG.md

@@ -6,6 +6,33 @@ This project uses GitHub Releases for detailed generated release notes. This fil
 
 ## Unreleased
 
+## 0.14.0 - 2026-07-03
+
+### Added
+
+- Adopted Pyright static type checking (basic mode) across the codebase.
+  Pyright runs in pre-commit and CI, catching type errors before runtime. The
+  `[tool.pyright]` configuration lives in `pyproject.toml`. Run locally with
+  `uv run pyright`.
+- Added support for Anthropic Claude Fable 5 (`claude-fable-5`) and Claude Sonnet 5
+  (`claude-sonnet-5`): 1M-token context, 128K max output, vision input, and
+  adaptive thinking with `low`/`medium`/`high`/`xhigh`/`max` effort levels. Both
+  are selectable in the Settings UI, `/model` picker, CLI flags, and env vars.
+- Added a `/goal` slash command that sets an autonomous completion condition the
+  agent works toward. After each turn, a read-only investigation sub-agent
+  verifies whether the goal is met; if not, the agent is nudged to continue
+  automatically until the goal is met, a turn cap is hit, or the user
+  pauses/cancels. Also available as `kolega-code ask --goal <condition>` for
+  run-to-completion from the CLI. Goal state persists with the session.
+
+### Fixed
+
+- Handled CRLF line endings in the edit, multi_edit, and write tools so diffs
+  and patches apply correctly on Windows-originated files.
+- Showed full diffs on the session changes screen instead of a capped preview.
+- Kept the turn worker alive during goal verification so the agent does not
+  stall while checking autonomous goal conditions.
+
 ## 0.13.0 - 2026-07-02
 
 ### Added

{kolega_code-0.13.0 → kolega_code-0.14.0}/CONTRIBUTING.md

@@ -57,12 +57,13 @@ to `.githooks`:
 uv run pre-commit run --all-files --show-diff-on-failure
 ```
 
-CI runs Ruff through pre-commit. You can also run Ruff directly for one-off linting and
-formatting checks:
+CI runs Ruff and Pyright through pre-commit. You can also run them directly for
+one-off checks:
 
 ```bash
 uv run ruff check .
 uv run ruff format --check .
+uv run pyright
 ```
 
 To apply formatting locally:
@@ -71,6 +72,27 @@ To apply formatting locally:
 uv run ruff format .
 ```
 
+### Type checking
+
+The project uses [Pyright](https://github.com/microsoft/pyright) in basic mode to
+catch type errors before runtime. The configuration lives in `[tool.pyright]` in
+`pyproject.toml`.
+
+Run the type checker locally:
+
+```bash
+uv run pyright
+```
+
+Guidelines for type annotations:
+
+- All new code should have type annotations on function signatures and class
+  attributes.
+- Use `# pyright: ignore[<rule>]` sparingly and only for third-party stub
+  limitations — never to silence errors in first-party code.
+- Prefer narrowing `Optional` values with `assert` or `if` guards over ignoring
+  the error.
+
 Run the fast test suite before opening a pull request:
 
 ```bash

{kolega_code-0.13.0 → kolega_code-0.14.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kolega-code
-Version: 0.13.0
+Version: 0.14.0
 Summary: Multi-agent coding in the terminal
 Project-URL: Homepage, https://kolega-ai.github.io/kolega-code/
 Project-URL: Documentation, https://kolega-ai.github.io/kolega-code/
@@ -255,6 +255,7 @@ Provides-Extra: cli
 Provides-Extra: dev
 Requires-Dist: genbadge[coverage]==1.1.3; extra == 'dev'
 Requires-Dist: pre-commit==4.6.0; extra == 'dev'
+Requires-Dist: pyright==1.1.411; extra == 'dev'
 Requires-Dist: pytest-asyncio==1.3.0; extra == 'dev'
 Requires-Dist: pytest-cov==7.0.0; extra == 'dev'
 Requires-Dist: pytest==9.0.3; extra == 'dev'

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/astro.config.mjs

@@ -23,8 +23,8 @@ export default defineConfig({
       // Brand fonts (Geist) + the Kolega Code theme. Order matters: fonts first
       // so the brand stylesheet can reference their family names.
       customCss: [
-        "@fontsource-variable/geist/index.css",
-        "@fontsource-variable/geist-mono/index.css",
+        "@fontsource-variable/ibm-plex-sans/index.css",
+        "@fontsource-variable/jetbrains-mono/index.css",
         "./src/styles/brand.css",
       ],
       social: [
@@ -105,6 +105,10 @@ export default defineConfig({
           label: "Gigacode",
           items: [{ label: "Workflow Orchestration", slug: "gigacode" }],
         },
+        {
+          label: "Autonomous Goals",
+          items: [{ label: "Goal-Conditioned Work", slug: "goal" }],
+        },
         {
           label: "Troubleshooting",
           items: [

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/package-lock.json

@@ -10,8 +10,8 @@
       "license": "Apache-2.0",
       "dependencies": {
         "@astrojs/starlight": "^0.40.0",
-        "@fontsource-variable/geist": "^5.2.9",
-        "@fontsource-variable/geist-mono": "^5.2.8",
+        "@fontsource-variable/ibm-plex-sans": "^5.2.8",
+        "@fontsource-variable/jetbrains-mono": "^5.2.8",
         "astro": "^6.4.7",
         "sharp": "^0.34.0"
       }
@@ -747,19 +747,19 @@
         "@expressive-code/core": "^0.43.1"
       }
     },
-    "node_modules/@fontsource-variable/geist": {
-      "version": "5.2.9",
-      "resolved": "https://registry.npmjs.org/@fontsource-variable/geist/-/geist-5.2.9.tgz",
-      "integrity": "sha512-TP+QSBG3wxKGPE33CbMy/L0Nu3qvJ6Fy81Yc4LnQ95xH+i+cfEp8fyU8/kfV14YwszxIFPhnoMTbjL71waVpyQ==",
+    "node_modules/@fontsource-variable/ibm-plex-sans": {
+      "version": "5.2.8",
+      "resolved": "https://registry.npmjs.org/@fontsource-variable/ibm-plex-sans/-/ibm-plex-sans-5.2.8.tgz",
+      "integrity": "sha512-n5PF2iFa0CZT0QYTPzxvZ39opC9LnU0zdoRccoADbs+Dtsd+lbXOZF7RNuIPHcQX1dKjF63sxnRImQIB5eD0Ag==",
       "license": "OFL-1.1",
       "funding": {
         "url": "https://github.com/sponsors/ayuhito"
       }
     },
-    "node_modules/@fontsource-variable/geist-mono": {
+    "node_modules/@fontsource-variable/jetbrains-mono": {
       "version": "5.2.8",
-      "resolved": "https://registry.npmjs.org/@fontsource-variable/geist-mono/-/geist-mono-5.2.8.tgz",
-      "integrity": "sha512-KI5bj+hkkRiHttYHmccotUZ80ZuZyai+RwI1d7UId0clkx/jXxlo8qYK8j54WzmpBjtMoEMPyllV7faDcj+6RA==",
+      "resolved": "https://registry.npmjs.org/@fontsource-variable/jetbrains-mono/-/jetbrains-mono-5.2.8.tgz",
+      "integrity": "sha512-WBA9elru6Jdp5df2mES55wuOO0WIrn3kpXnI4+W2ek5u3ZgLS9XS4gmIlcQhiZOWEKl95meYdvK7xI+ETLCq/Q==",
       "license": "OFL-1.1",
       "funding": {
         "url": "https://github.com/sponsors/ayuhito"

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/package.json

@@ -13,8 +13,8 @@
   },
   "dependencies": {
     "@astrojs/starlight": "^0.40.0",
-    "@fontsource-variable/geist": "^5.2.9",
-    "@fontsource-variable/geist-mono": "^5.2.8",
+    "@fontsource-variable/ibm-plex-sans": "^5.2.8",
+    "@fontsource-variable/jetbrains-mono": "^5.2.8",
     "astro": "^6.4.7",
     "sharp": "^0.34.0"
   },

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/public/favicon.svg

@@ -1,5 +1,5 @@
 <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24">
-  <rect width="24" height="24" rx="5" ry="5" fill="#c5ee21"/>
+  <rect width="24" height="24" rx="5" ry="5" fill="#39c5cf"/>
   <path d="M7 8L3 12L7 16" stroke="#000000" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" fill="none"/>
   <path d="M17 8L21 12L17 16" stroke="#000000" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" fill="none"/>
   <path d="M10 20L14 4" stroke="#000000" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" fill="none"/>

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/src/assets/coverage.svg

@@ -1 +1 @@
-<svg xmlns="http://www.w3.org/2000/svg" width="114" height="20" role="img" aria-label="coverage: 80.14%"><title>coverage: 80.14%</title><filter id="blur"><feGaussianBlur stdDeviation="16"/></filter><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="114" height="20" rx="3"/></clipPath><g clip-path="url(#r)"><rect width="61" height="20" fill="#555"/><rect x="61" width="53" height="20" fill="#67ac09"/><rect width="114" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><g transform="scale(.1)"><g aria-hidden="true" fill="#010101"><text x="315" y="150" fill-opacity=".8" filter="url(#blur)" textLength="510">coverage</text><text x="315" y="150" fill-opacity=".3" textLength="510">coverage</text></g><text x="315" y="140" textLength="510">coverage</text></g><g transform="scale(.1)"><g aria-hidden="true" fill="#010101"><text x="865" y="150" fill-opacity=".8" filter="url(#blur)" textLength="430">80.14%</text><text x="865" y="150" fill-opacity=".3" textLength="430">80.14%</text></g><text x="865" y="140" textLength="430">80.14%</text></g></g></svg>
+<svg xmlns="http://www.w3.org/2000/svg" width="114" height="20" role="img" aria-label="coverage: 80.35%"><title>coverage: 80.35%</title><filter id="blur"><feGaussianBlur stdDeviation="16"/></filter><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="114" height="20" rx="3"/></clipPath><g clip-path="url(#r)"><rect width="61" height="20" fill="#555"/><rect x="61" width="53" height="20" fill="#67ac09"/><rect width="114" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><g transform="scale(.1)"><g aria-hidden="true" fill="#010101"><text x="315" y="150" fill-opacity=".8" filter="url(#blur)" textLength="510">coverage</text><text x="315" y="150" fill-opacity=".3" textLength="510">coverage</text></g><text x="315" y="140" textLength="510">coverage</text></g><g transform="scale(.1)"><g aria-hidden="true" fill="#010101"><text x="865" y="150" fill-opacity=".8" filter="url(#blur)" textLength="430">80.35%</text><text x="865" y="150" fill-opacity=".3" textLength="430">80.35%</text></g><text x="865" y="140" textLength="430">80.35%</text></g></g></svg>

kolega_code-0.14.0/docs/src/assets/kolega-dark.svg

@@ -0,0 +1,15 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 290 30" width="290" height="30">
+  <defs>
+    <linearGradient id="g" x1="0" y1="0" x2="1" y2="0">
+      <stop offset="0%" stop-color="#39C5CF"/>
+      <stop offset="100%" stop-color="#9B7DD4"/>
+    </linearGradient>
+  </defs>
+  <text x="0" y="5" font-family="'JetBrains Mono','Cascadia Code','Fira Code',Menlo,Consolas,monospace" font-size="5" fill="url(#g)" font-weight="700" xml:space="preserve">
+    <tspan x="0" dy="0"> ██  ██   ██████  ██      ███████  ██████   █████    ██████  ██████  ██████  ███████</tspan>
+    <tspan x="0" dy="5.5"> ██ ██   ██    ██ ██      ██      ██       ██   ██  ██      ██    ██ ██   ██ ██</tspan>
+    <tspan x="0" dy="5.5"> ████    ██    ██ ██      █████   ██   ███ ███████  ██      ██    ██ ██   ██ █████</tspan>
+    <tspan x="0" dy="5.5"> ██ ██   ██    ██ ██      ██      ██    ██ ██   ██  ██      ██    ██ ██   ██ ██</tspan>
+    <tspan x="0" dy="5.5"> ██  ██   ██████  ███████ ███████  ██████  ██   ██   ██████  ██████  ██████  ███████</tspan>
+  </text>
+</svg>

kolega_code-0.14.0/docs/src/assets/kolega-light.svg

@@ -0,0 +1,15 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 290 30" width="290" height="30">
+  <defs>
+    <linearGradient id="g" x1="0" y1="0" x2="1" y2="0">
+      <stop offset="0%" stop-color="#0d7377"/>
+      <stop offset="100%" stop-color="#5a3d7a"/>
+    </linearGradient>
+  </defs>
+  <text x="0" y="5" font-family="'JetBrains Mono','Cascadia Code','Fira Code',Menlo,Consolas,monospace" font-size="5" fill="url(#g)" font-weight="700" xml:space="preserve">
+    <tspan x="0" dy="0"> ██  ██   ██████  ██      ███████  ██████   █████    ██████  ██████  ██████  ███████</tspan>
+    <tspan x="0" dy="5.5"> ██ ██   ██    ██ ██      ██      ██       ██   ██  ██      ██    ██ ██   ██ ██</tspan>
+    <tspan x="0" dy="5.5"> ████    ██    ██ ██      █████   ██   ███ ███████  ██      ██    ██ ██   ██ █████</tspan>
+    <tspan x="0" dy="5.5"> ██ ██   ██    ██ ██      ██      ██    ██ ██   ██  ██      ██    ██ ██   ██ ██</tspan>
+    <tspan x="0" dy="5.5"> ██  ██   ██████  ███████ ███████  ██████  ██   ██   ██████  ██████  ██████  ███████</tspan>
+  </text>
+</svg>

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/src/content/docs/cli/ask.md

@@ -15,8 +15,10 @@ kolega-code ask "<prompt>" [options]
 
 | Argument / option | Description |
 | --- | --- |
-| `prompt` | The prompt to send (required, positional) |
+| `prompt` | The prompt to send (optional when `--goal` is given; otherwise required) |
 | `--project <PATH>` | Project directory to work in (default `.`) |
+| `--goal <condition>` | Set an autonomous completion goal and loop until it is met or capped (no prompt required) |
+| `--goal-max-turns <N>` | Maximum evaluation turns before an unmet goal gives up (default 50) |
 | `--save` | Persist the session after the prompt completes |
 | `--json` | Emit response chunks and events as JSON |
 | `--browser-visible` | Launch visible Playwright browser windows |
@@ -77,6 +79,22 @@ it against the project's [Agent Skills](../../skills/):
 - `kolega-code ask "/my-skill do the thing"` activates the skill and runs the
   remaining prompt.
 
+## Goal mode
+
+Pass `--goal "<condition>"` to set an autonomous completion goal. The agent works
+toward the goal, and after each turn a read-only verifier checks whether it's met.
+The loop continues until the goal is met or the turn cap (default 50, override
+with `--goal-max-turns`) is reached. The positional `prompt` is optional with
+`--goal` — the CLI synthesizes the first work-turn message from the condition:
+
+```bash
+kolega-code ask --goal "all tests pass and ruff is clean" --project .
+kolega-code ask "start by fixing the parser" --goal "all tests pass" --project .
+```
+
+See [Goal-Conditioned Work](../../goal/) for the loop behavior, safety model, and
+JSON event details.
+
 ## JSON output
 
 With `--json`, the command streams newline-delimited JSON objects, each tagged with
@@ -93,6 +111,8 @@ The stream includes:
 | `chunk` | A streamed piece of the response |
 | `event` | An agent event (sub-agent activity, tool calls, terminal output) |
 | `skill` | Skill activation metadata |
+| `goal_eval` | Emitted after each goal evaluation (with `--goal`): `met`, `turns`, `reason` |
+| `goal_result` | Final goal outcome (with `--goal`): `met`, `turns`, `reason` |
 | `summary` | A final object with the chunk count and `session_id` |
 
 In plain (non-JSON) mode, the answer is written to **stdout** while sub-agent and
@@ -105,3 +125,12 @@ answer.
 confirmations. If you pass `--permission-mode ask`, shell commands and file edits
 prompt on stderr when stdin is interactive. Persisted allow rules are stored in
 the project at `.kolega/permissions.json`.
+
+## Exit codes
+
+| Code | Meaning |
+| --- | --- |
+| `0` | Success (or: the goal was met when using `--goal`) |
+| `1` | With `--goal`: the turn cap was reached without meeting the goal |
+| `2` | Configuration / usage error (e.g. invalid provider, missing API key) |
+| `130` | Interrupted (`Ctrl+C`) |

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/src/content/docs/cli/overview.md

@@ -78,5 +78,6 @@ in [Settings & API Keys](../../configuration/settings-and-api-keys/).
 | Code | Meaning |
 | --- | --- |
 | `0` | Success |
+| `1` | Goal not met: `ask --goal` reached the turn cap without completing the goal |
 | `2` | Configuration / usage error (e.g. invalid provider, missing API key, project path not found, Textual not installed) |
 | `130` | Interrupted (`Ctrl+C`) |

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/src/content/docs/configuration/providers-and-models.mdx

@@ -171,6 +171,8 @@ models in the TUI.
 
 | Provider/model | Values | Kolega Code default |
 | --- | --- | --- |
+| Anthropic `claude-fable-5` | `low`, `medium`, `high`, `xhigh`, `max` | `medium` |
+| Anthropic `claude-sonnet-5` | `low`, `medium`, `high`, `xhigh`, `max` | `medium` |
 | Anthropic `claude-opus-4-7` | `low`, `medium`, `high`, `xhigh`, `max` | `medium` |
 | DeepSeek `deepseek-v4-pro` | `none`, `high`, `max` | `high` |
 | Moonshot `kimi-k2.7-code` | `auto` | `auto` |

kolega_code-0.14.0/docs/src/content/docs/goal.md

@@ -0,0 +1,167 @@
+---
+title: Goal-Conditioned Work
+description: Set an autonomous completion goal the agent works toward until it is verifiably met.
+---
+
+Set a verifiable completion condition and the agent works autonomously toward it,
+verifying its own progress after every turn, until the goal is met, a turn cap is
+hit, or you stop it. It's available as the `/goal` slash command in the
+[TUI](../tui/interface/) and as `kolega-code ask --goal` from the
+[CLI](../cli/ask/).
+
+## How it works
+
+The goal loop runs after each completed work turn:
+
+1. **Work turn** — the agent uses its full toolset (read, edit, run commands, run
+   tests, dispatch sub-agents) to make progress toward the goal.
+2. **Verification** — a **read-only investigation sub-agent** inspects the current
+   codebase state and decides whether the goal is met, ending its reply with a
+   JSON verdict: `{"ok": true}` or `{"ok": false, "reason": "<remaining gap>"}`.
+3. **Met** — the goal completes, a confirmation message is shown, and the
+   active-goal prompt extension is dropped so subsequent turns are normal again.
+4. **Not met** — the agent is nudged to continue. The nudge includes the
+   verifier's stated remaining gap and the number of turns left before the cap, so
+   the agent can prioritize accordingly. The loop repeats from step 1.
+
+A **turn cap** (default **50**) is a safety backstop. If the goal isn't met within
+that many evaluation turns, the loop pauses and tells you to refine the goal or
+clear it.
+
+## Safety model
+
+:::note
+The verifier is a **read-only investigation sub-agent**. It can read files, search
+the code, and run commands or tests, but it **cannot edit** — so it can never game
+an autonomous goal by falsifying its own completion.
+:::
+
+A few guarantees that keep the loop safe:
+
+- The verifier inspects the codebase **fresh each evaluation** (stateless across
+  turns), so it judges the real current state, not a stale snapshot.
+- A malformed, unparseable, or failed verdict is **always treated as not-met**. A
+  broken evaluator can never falsely complete a goal — the loop keeps running or
+  hits the cap.
+- The verifier runs on the configured long-context model, separate from the
+  working agent's turn.
+
+## In the TUI: `/goal`
+
+Type `/goal` in the [composer](../tui/composer/) to set, check, or clear a goal.
+
+| Form | Effect |
+| --- | --- |
+| `/goal <condition>` | Set a goal and start working toward it |
+| `/goal` | Show goal status (condition, runtime, turns evaluated, tokens spent, verifier's latest reason) |
+| `/goal clear` | Remove the goal (aliases: `stop`, `off`, `reset`, `none`, `cancel`) |
+| `/goal -p <condition>` | Run-to-completion mode (alias: `--print`) — no pauses until the goal is met or capped |
+
+A few things to know:
+
+- **Esc pauses** the goal loop after the current turn finishes. Sending any
+  message **resumes** it from where it left off.
+- You can't set or clear a goal while a turn is active — stop the turn first.
+- Setting a new goal replaces an active unmet one.
+- When the turn cap is reached, the goal is **paused** with a note. Refine it with
+  `/goal <condition>` or remove it with `/goal clear`.
+- The **status dashboard** (Status side-panel tab) shows a `Goal` line with the
+  condition (truncated if long) and a state label — `active`, `paused`, or `met`.
+- **Persistence**: goal state is saved with the session and restored on resume, so
+  the loop picks up where it left off. See
+  [Sessions & Resuming](../tui/sessions-and-resume/).
+
+```text
+/goal all tests pass and the linter is clean
+```
+
+## From the CLI: `ask --goal`
+
+```bash
+kolega-code ask --goal "<condition>" [--goal-max-turns N] [options]
+```
+
+The positional `prompt` is **optional** when `--goal` is given — the CLI
+synthesizes the first work-turn message from the condition. You can still pass a
+prompt to give the agent a head start:
+
+```bash
+# No prompt needed — the condition drives the loop
+kolega-code ask --goal "all tests pass and ruff is clean" --project .
+
+# Lower the turn cap for a bounded task
+kolega-code ask --goal "the failing test in test_parser.py passes" --goal-max-turns 10 --json
+
+# Give the agent a starting point, then let the goal loop take over
+kolega-code ask "start by fixing the parser" --goal "all tests pass" --project .
+```
+
+| Option | Description |
+| --- | --- |
+| `--goal <condition>` | Set an autonomous completion goal and loop until it is met or capped (no prompt required) |
+| `--goal-max-turns <N>` | Maximum evaluation turns before an unmet goal gives up (default 50) |
+
+### Plain output
+
+In plain (non-JSON) mode, after each evaluation the CLI prints a line to
+**stderr**:
+
+```text
+[goal] turn 1: not met — test_parser.py still fails on test_parse_empty
+[goal] turn 2: not met — two tests still failing in test_parser.py
+```
+
+When the loop ends, a final summary line is printed:
+
+```text
+[goal] met after 3 turn(s)
+# or
+[goal] not met (turn cap reached) after 10 turn(s)
+```
+
+The agent's response text is written to **stdout** as usual, so piping stdout still
+gives you just the answer.
+
+### Exit code
+
+With `--goal`, the exit code reflects the outcome:
+
+| Code | Meaning |
+| --- | --- |
+| `0` | The goal was met |
+| `1` | The turn cap was reached without meeting the goal |
+
+## JSON output
+
+With `--json`, the command streams newline-delimited JSON objects as usual, plus
+two goal-specific `kind` values alongside the existing `chunk`, `event`, and
+`summary` kinds:
+
+| `kind` | Meaning |
+| --- | --- |
+| `goal_eval` | Emitted after each evaluation: `{met, turns, reason}` |
+| `goal_result` | Final outcome: `{met, turns, reason}` |
+
+Example stream (abbreviated):
+
+```json
+{"kind": "chunk", "data": {"type": "response", "content": "Fixing the parser…"}}
+{"kind": "goal_eval", "data": {"met": false, "turns": 1, "reason": "test_parse_empty still fails"}}
+{"kind": "chunk", "data": {"type": "response", "content": "Added the empty-input guard…"}}
+{"kind": "goal_eval", "data": {"met": true, "turns": 2, "reason": ""}}
+{"kind": "goal_result", "data": {"met": true, "turns": 2, "reason": ""}}
+{"kind": "summary", "chunks": 4, "session_id": "abc123"}
+```
+
+## Tips
+
+- Write conditions that are **verifiable** — "all tests pass", "the file exists",
+  "the command succeeds" — rather than subjective ones like "make the code better".
+- Keep the turn cap in mind for open-ended goals. Lower it with `--goal-max-turns`
+  for bounded tasks where you don't want the agent to spin for 50 turns.
+- The verifier is strict. "Mostly done" is not met — the goal is met only when
+  there is concrete evidence (described files/code exist, relevant tests or
+  commands pass).
+- A good condition names a **checkable outcome**, not a process. "Refactor the
+  parser" is a process; "the parser module has 100% test coverage and all tests
+  pass" is a checkable outcome.

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/src/content/docs/tui/interface.md

@@ -27,7 +27,7 @@ appears above the composer until those prompts are sent. See
 
 | Tab | What it shows |
 | --- | --- |
-| **Status** | The active provider/model and thinking effort, the current interaction mode (Build/Plan), permission mode, the agent's turn state (idle, generating, thinking, running a tool, running sub-agents, waiting for input, …), token usage, and context warnings. |
+| **Status** | The active provider/model and thinking effort, the current interaction mode (Build/Plan), permission mode, the agent's turn state (idle, generating, thinking, running a tool, running sub-agents, waiting for input, …), token usage, context warnings, and the active [goal](../../goal/) status when a goal is set. |
 | **Logs** | Optional. Launch with `--show-logs` to show a timestamped, color-coded diagnostic activity log. New entries preserve manual scrollback and an indicator flags unseen entries when you're on another tab. |
 | **Terminal** | Live output from commands the agent runs. |
 | **Planning** | The current **Plan** (markdown from the planning agent) and the shared **Task List** that both modes can edit. |

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/src/content/docs/tui/sessions-and-resume.md

@@ -52,4 +52,6 @@ kolega-code sessions delete <session_id>
 :::note
 All CLI sessions use the CLI-specific coding-agent prompt, including resumed ones.
 Resuming restores the conversation history so the agent picks up with full context.
+An active [goal](../../goal/) is also saved with the session and restored on resume,
+so the autonomous loop continues where it left off.
 :::

{kolega_code-0.13.0 → kolega_code-0.14.0}/docs/src/content/docs/tui/slash-commands.md

@@ -43,6 +43,7 @@ These control the app and your session.
 | `/login` | Sign in to a provider, e.g. `/login chatgpt` |
 | `/logout` | Sign out of a provider, e.g. `/logout chatgpt` |
 | `/gigacode` | Toggle [gigacode](../../gigacode/) workflow orchestration on or off |
+| `/goal` | Set, show, or clear an autonomous completion goal |
 | `/queue-clear` | Clear queued follow-up messages |
 | `/copy` | Copy the last response to the clipboard |
 | `/diagnostics` | Show version, model/endpoint, and the local diagnostics log path |
@@ -52,6 +53,13 @@ These control the app and your session.
 | `/quit` | Save the session and exit |
 | `/exit` | Save the session and exit |
 
+Run `/goal <condition>` to set an autonomous completion goal the agent works
+toward, verifying its own progress after each turn until the goal is met, the turn
+cap is hit, or you pause it. `/goal` (no args) shows the goal status; `/goal clear`
+(aliases: `stop`, `off`, `reset`, `none`, `cancel`) removes it. See
+[Goal-Conditioned Work](../../goal/) for the full loop behavior, safety model, and
+examples.
+
 Run `/model` to open a selectable list of supported models for the current
 provider. You can also switch directly with `/model <name>`.