akm-cli 0.10.261012-beta → 0.10.261014-beta

package/dist/integrations/agent/index.js

@@ -2,6 +2,6 @@
 // License, v. 2.0. If a copy of the MPL was not distributed with this
 // file, You can obtain one at https://mozilla.org/MPL/2.0/.
 export { DEFAULT_AGENT_TIMEOUT_MS } from "./config.js";
-export { _setAgentDetectForTests, defaultWhich, detectAgentCliProfiles, pickDefaultAgentProfile } from "./detect.js";
+export { _setAgentDetectForTests, defaultWhich, detectAgentCliProfiles } from "./detect.js";
 export { getBuiltinAgentProfile, listBuiltinAgentProfiles, } from "./profiles.js";
 export { buildProposePrompt, buildReflectPrompt, parseAgentProposalPayload } from "./prompts.js";

package/dist/integrations/agent/profiles.js

@@ -2,19 +2,18 @@
 // License, v. 2.0. If a copy of the MPL was not distributed with this
 // file, You can obtain one at https://mozilla.org/MPL/2.0/.
 /**
- * Built-in profile registry for external agent CLIs (v1 spec §12.1).
+ * Built-in profile registry for external agent CLIs.
  *
  * A `AgentProfile` is the minimum metadata required to shell-out to a
- * coding-agent CLI. Named engines lower canonical harness metadata into this
+ * coding-agent CLI. A harness lowers canonical harness metadata into this
  * intentionally small internal shape. The wrapper is in `./spawn.ts`.
  */
-import { COMMON_SPAWN_ENV_PASSTHROUGH } from "../../core/spawn-env.js";
 import { HARNESS_REGISTRY } from "../harnesses/index.js";
 /**
  * Built-in profiles for the agent CLIs akm knows out of the box, DERIVED from
  * the harness registry: each harness that declares `profile` defaults gets one,
  * named by its id. The fields are conservative defaults, overridable from user
- * config. Engine lowering selects captured stdio for unattended dispatch.
+ * config. Request lowering selects captured stdio for unattended dispatch.
  *
  * Built lazily: the registry transitively imports this module, so it cannot be
  * read at module load.
@@ -25,39 +24,15 @@ function builtinProfiles() {
         return builtins;
     builtins = {};
     for (const harness of HARNESS_REGISTRY) {
-        if (!harness.profile)
-            continue;
         builtins[harness.id] = {
             name: harness.id,
             bin: harness.profile.bin,
             args: harness.profile.args,
             stdio: "interactive",
-            // AKM_EVENT_SOURCE (in the common set) carries usage-event provenance so akm
-            // invocations a spawned agent makes are recorded as machine traffic; it is a
-            // provenance tag, never a secret.
-            envPassthrough: [...COMMON_SPAWN_ENV_PASSTHROUGH, ...harness.profile.envPassthrough],
-            parseOutput: "text",
         };
     }
     return builtins;
 }
-/**
- * Binary the `opencode-sdk` harness needs on PATH.
- *
- * The embedded client is not self-contained, and that is the SDK's own design
- * rather than a consequence of how akm drives it: `@opencode-ai/sdk` (OpenCode 1)
- * ships with `"dependencies": {}` and its `createOpencodeServer` is itself a
- * `spawn("opencode", ["serve", ...])`; `@opencode/client` (OpenCode 2, the
- * default) is an HTTP client only and starts nothing. akm's runner spawns a
- * private `opencode serve` of the engine's detected major (
- * never the v2 background service) and talks HTTP to it through that major's
- * adapter (see `harnesses/opencode-sdk/sdk-runner.ts`, `v1-adapter.ts`, `v2-adapter.ts`),
- * so the `opencode` binary gates the SDK path exactly as it gates the CLI
- * path — a host with the npm package but no binary can dispatch neither.
- * `opencode-sdk` deliberately declares no `profile` — it dispatches
- * without argv construction — so this is the one place that pairing lives.
- */
-export const OPENCODE_SDK_SERVER_BIN = "opencode";
 /** Returns the built-in descriptor for a canonical harness id. */
 export function getBuiltinAgentProfile(name) {
     return builtinProfiles()[name];
@@ -73,3 +48,24 @@ export function listBuiltinAgentProfiles() {
     }
     return out;
 }
+/** The profile a harness runner spawns: its platform's built-in defaults under what the config set. */
+export function harnessProfile(runner) {
+    return {
+        name: runner.name,
+        platform: runner.platform,
+        bin: runner.bin,
+        args: runner.args,
+        stdio: "captured",
+        ...(runner.model ? { model: runner.model } : {}),
+    };
+}
+/** The profile of akm's own runtime, the pinned OpenCode process that runs agent work over a connection. It names the connection; the program is fetched and started by `harnesses/opencode/runtime/`. */
+export function runtimeProfile(connection) {
+    return {
+        name: connection.name,
+        bin: "",
+        args: [],
+        stdio: "captured",
+        model: connection.model,
+    };
+}

package/dist/integrations/agent/prompts.js

@@ -4,14 +4,14 @@
 /**
  * Shared prompt builders for proposal-producing agent commands (#226).
  *
- * `akm reflect` and `akm proposal new` both ask the configured engine for a
+ * `akm reflect` and `akm proposal new` both ask the connection or harness they run on for a
  * structured proposal payload. The prompts are intentionally similar and share
- * their construction here. `proposal new` asks every engine kind for the JSON
+ * their construction here. `proposal new` asks a connection and a harness alike for the JSON
  * object below, with {@link PROPOSAL_JSON_SCHEMA} as the request's output
- * schema. Reflect asks every engine kind for its own contract, JSON Schema or
+ * schema. Reflect asks both for its own contract, JSON Schema or
  * framed markdown ({@link ReflectOutputMode}).
  *
- * The output an engine must produce for `proposal new` is a strict JSON object:
+ * The output `proposal new` needs back is a strict JSON object:
  *
  * ```json
  * {
@@ -42,7 +42,7 @@ const TYPE_HINTS = {
     lesson: "lesson assets MUST start with frontmatter containing `description` and `when_to_use` keys (both non-empty). Body: 1–3 short paragraphs of practical guidance. A lesson is NOT a restatement of the source asset — it answers: When should I reach for this? What goes wrong without it? What did real use reveal that the asset itself doesn't say?",
     skill: "skill assets are stored as `skills/<name>/SKILL.md`. Frontmatter typically includes `name`, `description`, and `when_to_use`.",
     command: "command assets are markdown with optional frontmatter (`name`, `description`). The body is the prompt template the user invokes.",
-    agent: "agent assets are markdown with frontmatter describing the agent role (`name`, `description`, optional `tools`, `model`).",
+    agent: "agent assets are markdown with frontmatter describing the agent role (`name`, `description`, optional `model`).",
     knowledge: "knowledge assets are reference markdown documents. Include a top-level `# Title` and concise sections.",
     memory: "memory assets are short factual notes the user wants persisted across sessions. Frontmatter usually includes `description`.",
     workflow: "workflow assets are markdown describing a multi-step process. Include `# <Title>` and ordered `## Step N` sections.",
@@ -75,7 +75,7 @@ export const REFLECT_CONTENT_CAP = 12_000;
  */
 export const REFLECT_TRUNCATION_MARKER = "... [truncated — focus on the visible portion]";
 /**
- * The JSON envelope `proposal new` asks the engine to honour. The wrapper code
+ * The JSON envelope `proposal new` asks the model to honour. The wrapper code
  * uses `JSON.parse(stdout)` to extract the payload — anything outside the JSON
  * object will be treated as a parse error.
  */
@@ -96,10 +96,10 @@ const RESPONSE_CONTRACT_JSON = [
 ].join("\n");
 /**
  * The JSON Schema of {@link RESPONSE_CONTRACT_JSON}'s object, the output
- * schema `proposal new` sends with its request: an LLM engine gets it as
- * `response_format`, codex as `--output-schema`, every agent engine as the
- * shared schema instruction. It is in the strict form those native channels
- * accept (every property required, no others), so it leaves out the optional
+ * schema `proposal new` sends with its request: a connection gets it as
+ * `response_format`, a harness as the shared schema instruction. It
+ * is in the strict form a provider's native channel accepts (every property
+ * required, no others), so it leaves out the optional
  * `frontmatter`, which `content` already carries. The reply is held only to
  * what {@link validateProposalPayload} needs.
  */
@@ -252,12 +252,12 @@ export function buildReflectPrompt(input) {
     }
     if (input.assetContent?.trim()) {
         // Cap defaults to REFLECT_CONTENT_CAP (12 000 chars) to stay well under OS
-        // ARG_MAX when the prompt is passed as a CLI argument to opencode/claude
-        // — large assets (wiki snapshots, long runbooks) would otherwise trigger
-        // E2BIG on posix_spawn. Direct-LLM callers never touch argv, so they may
-        // pass a larger, context-aware `contentBudgetChars` (#952); it is never
-        // allowed below the flat floor, which the caller enforces before calling in.
-        const contentCap = input.contentBudgetChars ?? REFLECT_CONTENT_CAP;
+        // ARG_MAX when the prompt is passed as a CLI argument — large assets (wiki
+        // snapshots, long runbooks) would otherwise trigger E2BIG on posix_spawn.
+        // A prompt on stdin or over HTTP never touches argv, so the caller passes a
+        // larger, context-aware `contentBudgetChars` (#952), or `null` for none; a
+        // number is never allowed below the flat floor, which the caller enforces.
+        const contentCap = input.contentBudgetChars === null ? Number.POSITIVE_INFINITY : (input.contentBudgetChars ?? REFLECT_CONTENT_CAP);
         const body = input.assetContent.trimEnd();
         const truncated = body.length > contentCap;
         sections.push(truncated
@@ -297,19 +297,6 @@ export function buildReflectPrompt(input) {
     if (input.avoidPatterns && input.avoidPatterns.length > 0) {
         sections.push(`## Avoid These Patterns\nRun-only guidance: do not copy this heading, explanation, or any diagnostic below into the proposed asset.\nPrevious assets in this run produced these errors — do not repeat them:\n${input.avoidPatterns.map((e) => `- ${e}`).join("\n")}`);
     }
-    // R-1 / #372: Self-Refine (arXiv:2303.17651) — inject prior draft as critique target.
-    // On refinement iterations (iter > 0), the agent is shown its previous proposal
-    // and asked to self-critique and improve it rather than starting from scratch.
-    if (input.priorDraft?.trim()) {
-        sections.push("## Self-Refine: Critique and Improve\n" +
-            "The following is your previous draft proposal. " +
-            "Identify specific weaknesses: missing evidence, vague wording, incomplete frontmatter, " +
-            "or claims that duplicate existing content without adding new signal. " +
-            "Then produce an improved version that addresses those weaknesses.\n\n" +
-            "Previous draft:\n```\n" +
-            input.priorDraft.trimEnd() +
-            "\n```");
-    }
     if (input.ref && input.assetContent?.trim()) {
         const problems = frontmatterProblems(input.assetContent);
         sections.push(problems.length > 0
@@ -321,7 +308,7 @@ export function buildReflectPrompt(input) {
 }
 /**
  * Build the prompt for `akm proposal new <type> <name> --task ...`. Asks the
- * engine to author a brand-new asset of the given type fulfilling `task`, and
+ * model to author a brand-new asset of the given type fulfilling `task`, and
  * to return it as the JSON object of {@link RESPONSE_CONTRACT_JSON}.
  */
 export function buildProposePrompt(input) {

package/dist/integrations/agent/request-lowering.js

@@ -3,25 +3,8 @@
 // file, You can obtain one at https://mozilla.org/MPL/2.0/.
 import { ConfigError } from "../../core/errors.js";
 import { withSchemaInstruction } from "../../core/structured.js";
-import { MODEL_WORK_POLICY_ID } from "../../execution/source.js";
-import { getHarness } from "../harnesses/index.js";
 import { composeConversationFallbackPrompt } from "./conversation-fallback.js";
 import { composePersonaFallbackPrompt } from "./persona-fallback.js";
-/** A tool selection that actually names tools (not omitted, null, or empty). */
-export function hasToolSelection(value) {
-    if (value === null || value === undefined)
-        return false;
-    if (typeof value === "string" || Array.isArray(value))
-        return value.length > 0;
-    return Object.keys(value).length > 0;
-}
-function translatesTools(mode, value) {
-    if (mode === "all")
-        return true;
-    if (typeof value === "string" || Array.isArray(value))
-        return mode === "flat" || mode === "sdk";
-    return mode === "sdk" && value !== null && Object.values(value).every((entry) => typeof entry === "boolean");
-}
 /** The notice for a request field a transport does not carry; dispatch continues. */
 export function untranslated(adapter, field) {
     return {
@@ -32,94 +15,77 @@ export function untranslated(adapter, field) {
         message: `The ${adapter} lowerer does not translate resolved field ${field}; dispatch will continue optimistically.`,
     };
 }
-/** Request fields carrying adapter-owned extensions, which no transport translates. */
-export function extensionFields(request) {
-    return [
-        ...(request.extensions ? ["extensions"] : []),
-        ...(request.command.extensions ? ["command.extensions"] : []),
-        ...(request.persona?.extensions ? ["persona.extensions"] : []),
-        ...(request.engine.extensions ? ["engine.extensions"] : []),
-        ...(request.runtime.extensions ? ["runtime.extensions"] : []),
-    ];
-}
 /**
- * Build one harness's request builder: map a resolved request onto the
- * harness-neutral {@link AgentDispatchRequest} its argv builder consumes,
- * composing into the prompt what the harness has no channel for.
+ * Lower a resolved request for one target (a harness, or akm's own runtime): map
+ * it onto the harness-neutral {@link AgentDispatchRequest} its builder consumes,
+ * composing into the prompt what the target has no channel for.
  */
-export function createAgentRequestLowerer(options) {
-    const supportedInference = new Set(options.inference ?? []);
-    return (_profile, request) => {
-        const modelWork = request.authorization.policy?.id === MODEL_WORK_POLICY_ID;
-        const notices = [];
-        const skip = (field) => {
-            notices.push(untranslated(options.adapter, field));
-        };
-        let prompt = request.command.content;
-        if (request.conversation && request.conversation.length > 0) {
-            prompt = composeConversationFallbackPrompt(request.conversation, prompt);
-            notices.push({
-                code: "conversation-prompt-composed",
-                severity: "warning",
-                adapter: options.adapter,
-                field: "conversation",
-                message: `The ${options.adapter} transport has no native multi-message channel; AKM composed the conversation prefix into one deterministic JSON prompt block.`,
-            });
-        }
-        const dispatch = { prompt };
-        if (request.persona) {
-            if (options.personaChannel === "native")
-                dispatch.systemPrompt = request.persona.content;
-            else {
-                const composed = composePersonaFallbackPrompt(request.persona.content, prompt, options.adapter);
-                prompt = composed.prompt;
-                notices.push(...composed.notices);
-            }
-        }
-        else if (typeof request.agent === "string") {
-            if (!getHarness(options.adapter)?.capabilities.nativeAgent) {
-                throw new ConfigError(`The ${options.adapter} transport cannot consume native agent selector ${JSON.stringify(request.agent)}.`, "INVALID_CONFIG_FILE");
-            }
-            dispatch.agent = request.agent;
-        }
-        // Every agent transport gets the schema as the one instruction; a harness
-        // with a native channel (codex --output-schema) also reads dispatch.schema.
-        if (request.outputSchema) {
-            prompt = withSchemaInstruction(prompt, request.outputSchema);
-            dispatch.schema = request.outputSchema;
+export function lowerRequest(target, request) {
+    const supportedInference = new Set(target.inference ?? []);
+    const adapter = target.id;
+    const notices = [];
+    const skip = (field) => {
+        notices.push(untranslated(adapter, field));
+    };
+    let prompt = request.command.content;
+    if (request.conversation && request.conversation.length > 0) {
+        prompt = composeConversationFallbackPrompt(request.conversation, prompt);
+        notices.push({
+            code: "conversation-prompt-composed",
+            severity: "warning",
+            adapter,
+            field: "conversation",
+            message: `The ${adapter} transport has no native multi-message channel; AKM composed the conversation prefix into one deterministic JSON prompt block.`,
+        });
+    }
+    const dispatch = { prompt };
+    if (request.persona) {
+        if (target.systemPrompt)
+            dispatch.systemPrompt = request.persona.content;
+        else {
+            const composed = composePersonaFallbackPrompt(request.persona.content, prompt, adapter);
+            prompt = composed.prompt;
+            notices.push(...composed.notices);
         }
-        dispatch.prompt = prompt;
-        if (request.model)
-            dispatch.model = request.model.resolved;
-        if (Object.hasOwn(request, "inference")) {
-            dispatch.inference = request.inference ?? null;
-            for (const key of Object.keys(request.inference ?? {}).sort()) {
-                if (!(modelWork && supportedInference.has(key)))
-                    skip(`inference.${key}`);
-            }
+    }
+    else if (typeof request.agent === "string" && request.modelWork) {
+        // Model work is not agent work: a native agent would replace the confined run.
+        notices.push({
+            code: "model-work-agent-ignored",
+            severity: "warning",
+            adapter,
+            field: "agent",
+            message: `The ${adapter} transport ignores native agent ${JSON.stringify(request.agent)} for model work.`,
+        });
+    }
+    else if (typeof request.agent === "string") {
+        if (!target.nativeAgent) {
+            throw new ConfigError(`The ${adapter} transport cannot consume native agent selector ${JSON.stringify(request.agent)}.`, "INVALID_CONFIG_FILE");
         }
-        if (modelWork) {
-            if (!getHarness(options.adapter)?.capabilities.modelWork) {
-                throw new ConfigError(`The ${options.adapter} transport cannot enforce the model-work tool policy.`, "INVALID_CONFIG_FILE");
-            }
-            // The builder selects its own confined agent or flags; another agent would replace them.
-            if (dispatch.agent) {
-                throw new ConfigError(`The ${options.adapter} transport cannot run native agent ${JSON.stringify(dispatch.agent)} under the model-work tool policy.`, "INVALID_CONFIG_FILE");
-            }
+        dispatch.agent = request.agent;
+    }
+    // Every agent transport gets the schema as the one instruction.
+    if (request.outputSchema)
+        prompt = withSchemaInstruction(prompt, request.outputSchema);
+    dispatch.prompt = prompt;
+    if (request.model)
+        dispatch.model = request.model.model;
+    for (const key of Object.keys(request.inference ?? {}).sort()) {
+        if (!supportedInference.has(key))
+            skip(`inference.${key}`);
+    }
+    if (request.modelWork) {
+        if (target.confinable)
             dispatch.modelWork = true;
+        else {
+            notices.push({
+                code: "model-work-unrestricted",
+                severity: "info",
+                adapter,
+                field: "modelWork",
+                message: `The ${adapter} transport cannot restrict its tools for model work; the run is unrestricted, in a scratch directory.`,
+            });
         }
-        else if (request.tools !== undefined) {
-            // An explicit empty selection still reaches the builder (e.g. an empty allowlist).
-            if (hasToolSelection(request.tools) && !translatesTools(options.tools, request.tools)) {
-                throw new ConfigError(`The ${options.adapter} transport cannot enforce the resolved tool policy.`, "INVALID_CONFIG_FILE");
-            }
-            dispatch.tools = request.tools;
-        }
-        const settings = request.runtime.settings;
-        if (settings && Object.keys(settings).length > 0)
-            skip("runtime.settings");
-        for (const field of extensionFields(request))
-            skip(field);
-        return { prompt, dispatch, notices };
-    };
+    }
+    return { dispatch, notices };
 }