experimental-ash 0.15.0 → 0.16.1

package/CHANGELOG.md

@@ -1,5 +1,21 @@
 # experimental-ash
 
+## 0.16.1
+
+### Patch Changes
+
+- 8cf749c: Expose the clicking actor as a `user` field on `SlackInteractionAction` so `onInteraction` handlers can attribute resolutions without re-parsing the raw payload.
+
+## 0.16.0
+
+### Minor Changes
+
+- 4d7af3a: Remove `SlackThread.threadId` from the public Slack channel API and clarify Slack continuation-token encoding. Slack callers should use `ctx.slack.channelId` and `ctx.slack.threadTs` for real Slack API identifiers, while Ash keeps continuation tokens as channel-local values before runtime namespacing.
+
+### Patch Changes
+
+- 8e4067a: Remove Ash-owned AI Gateway reporting tags from harness model calls. Authored `providerOptions.gateway.tags` and `mergeGatewayAutoCaching` behavior are unchanged.
+
 ## 0.15.0
 
 ### Minor Changes
@@ -88,7 +104,7 @@
 
   The resulting `ctx` exposes two handles so call sites read naturally:
 
-  - `ctx.thread` (`SlackThread`) — `post`, `postEphemeral`, `startTyping`, `refresh`, `recentMessages`, `threadId`, `mentionUser`. Reads as "post a reply to this thread".
+  - `ctx.thread` (`SlackThread`) — `post`, `postEphemeral`, `startTyping`, `refresh`, `recentMessages`, `mentionUser`. Reads as "post a reply to this thread".
   - `ctx.slack` (`SlackHandle`) — `channelId`, `threadTs`, `teamId`, `request(op, body)`, `uploadFiles(...)`. Reads as "raw Slack API, possibly not the bound thread".
 
   `ctx.thread.post(...)` accepts ergonomic bare forms (`string` → `{ markdown }`, `CardElement` → `{ card }`) in addition to the explicit `SlackPostInput` variants:

package/dist/docs/public/agent-ts.md

@@ -33,11 +33,13 @@ Use `agent.ts` for:
 
 - model selection
 - metadata you want preserved in runtime traces
-- human-in-the-loop policy
 - hosted-build packaging controls
 - compaction settings
 - provider-specific model options
 
+Per-tool approval (formerly "human-in-the-loop policy") lives on each tool via `needsApproval`,
+not in `agent.ts`. See [Human In The Loop](./human-in-the-loop.md).
+
 For OpenTelemetry configuration, use `instrumentation.ts` instead. See
 [`instrumentation.ts`](./instrumentation.md).
 
@@ -130,29 +132,10 @@ server output instead of being bundled.
 The shared per-run workspace is configured on the sandbox, not on `agent.ts`. See
 [Workspace](./workspace.md) and [Sandboxes](./sandbox.md).
 
-## `humanInTheLoop`
-
-`humanInTheLoop` controls whether Ash should pause for approval before executing model-visible
-tools.
-
-```ts
-export default defineAgent({
-  humanInTheLoop: {
-    mode: "require-approval",
-  },
-});
-```
-
-Supported fields:
-
-- `mode: "require-approval" | "auto-allow"`
-- `allow?: readonly string[]`
-- `ask?: readonly string[]`
-
-Use `require-approval` when the default should be "ask first". Use `auto-allow` when the default
-should be "run immediately" and only a few named tools should still require approval.
+## Human In The Loop
 
-See [Human In The Loop](./human-in-the-loop.md) for the runtime flow, `input.requested`
+Approval policy lives on individual tools via `needsApproval` in `defineTool({...})`, not on
+`agent.ts`. See [Human In The Loop](./human-in-the-loop.md) for the runtime flow, `input.requested`
 events, and HTTP response payloads.
 
 ## Telemetry

package/dist/docs/public/context-control.md

@@ -66,7 +66,6 @@ Use the weather tool before answering forecast or temperature questions.
 
 ```md
 ---
-name: research
 description: Research unfamiliar topics before answering with confidence.
 ---
 
@@ -84,7 +83,8 @@ prompt.
 ### Skill Rules That Matter
 
 - packaged `SKILL.md` files must include YAML frontmatter
-- `name` and `description` are required in packaged skill frontmatter
+- `description` is required in packaged skill frontmatter; identity is derived from the directory
+  name
 - flat markdown skills may omit frontmatter
 - tools stay visible whether or not a skill is activated
 

package/dist/docs/public/evals.md

@@ -39,9 +39,7 @@ import { defineEvalSuite } from "experimental-ash/evals";
 import { Run } from "experimental-ash/evals/scores";
 
 export default defineEvalSuite({
-  id: "weather-smoke",
   model: "openai/gpt-5.4-mini",
-  name: "Weather Smoke",
   description: "Basic message and tool-usage coverage for the weather agent.",
   cases: [
     {
@@ -55,13 +53,12 @@ export default defineEvalSuite({
 ```
 
 `defineEvalSuite(...)` validates the suite shape and returns a normalized suite object that the CLI
-can discover and run.
+can discover and run. Suite identity is derived from the `.eval.ts` file path — authored
+definitions do not carry an `id` or `name`.
 
 Every suite must provide:
 
-- `id`
 - `model`
-- `name`
 - `scores`
 - either `cases` or `load`
 
@@ -88,9 +85,7 @@ You can provide cases directly:
 
 ```ts
 export default defineEvalSuite({
-  id: "smoke",
   model: "openai/gpt-5.4-mini",
-  name: "Smoke",
   cases: [
     {
       id: "hello",
@@ -110,9 +105,7 @@ import { loadYaml } from "experimental-ash/evals/loaders";
 import { Text } from "experimental-ash/evals/scores";
 
 export default defineEvalSuite({
-  id: "marketing-sql",
   model: "openai/gpt-5.4-mini",
-  name: "Marketing SQL",
   async load() {
     const doc = await loadYaml("evals/data/cases.yaml");
     return (doc.evals as readonly { task: string; prompt: string; sql: string }[]).map((row) => ({
@@ -131,9 +124,7 @@ message.
 
 ```ts
 defineEvalSuite({
-  id: "weather-task",
   model: "openai/gpt-5.4-mini",
-  name: "Weather Task",
   task: {
     prompt: (testCase) => `Answer the user: ${String(testCase.input)}`,
     parseOutput: (result) => result.finalMessage,
@@ -149,9 +140,7 @@ Suite thresholds let you relax a scorer from the default exact-match requirement
 
 ```ts
 defineEvalSuite({
-  id: "weather",
   model: "openai/gpt-5.4-mini",
-  name: "Weather",
   cases: [{ id: "hello", input: "Hello", expected: "Hello" }],
   scores: [Run.didNotFail(), Text.includes()],
   thresholds: {
@@ -180,8 +169,6 @@ import { defineEvalSuite } from "experimental-ash/evals";
 import { Autoevals, Run } from "experimental-ash/evals/scores";
 
 export default defineEvalSuite({
-  id: "support-quality",
-  name: "Support Quality",
   model: "openai/gpt-5.4-mini",
   cases: [{ id: "refund", input: "How do I get a refund?", expected: "refund policy" }],
   scores: [Run.didNotFail(), Autoevals.factuality()],

package/dist/docs/public/project-layout.md

@@ -34,7 +34,7 @@ my-agent/
 | Path                                   | Purpose                                                                                                                                                                                                                                | Notes                                                                                                                                                                                                                                           |
 | -------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
 | `instructions.md` or `instructions.ts` | Base instructions prompt                                                                                                                                                                                                               | Exactly one instructions prompt is expected. Module-backed sources execute once at build time; the resulting markdown is baked into the compiled manifest. The legacy `system.{md,ts,...}` slot is still discovered with a deprecation warning. |
-| `agent.ts`                             | Runtime config                                                                                                                                                                                                                         | Model, metadata, workspace, build, compaction                                                                                                                                                                                                   |
+| `agent.ts`                             | Runtime config                                                                                                                                                                                                                         | Model, metadata, build, compaction, modelOptions                                                                                                                                                                                                |
 | `instrumentation.ts`                   | Telemetry config                                                                                                                                                                                                                       | OTel exporter setup and AI SDK span settings; auto-discovered and run before agent code                                                                                                                                                         |
 | `channels/`                            | HTTP or messaging entrypoints                                                                                                                                                                                                          | Root-only today                                                                                                                                                                                                                                 |
 | `connections/`                         | External service connections (MCP)                                                                                                                                                                                                     | Each file defines one connection; name derived from filename                                                                                                                                                                                    |

package/dist/docs/public/skills.md

@@ -45,7 +45,7 @@ The smallest skill is just a markdown file.
 Use the weather tool before answering forecast or temperature questions.
 ```
 
-For flat skills, Ash derives the skill id and name from the file path.
+For flat skills, Ash derives the skill id from the file path.
 
 You can also add frontmatter when you want to control the description explicitly:
 
@@ -63,7 +63,6 @@ Use a packaged skill when the procedure needs sibling files.
 
 ```md
 ---
-name: research
 description: Research unfamiliar topics before answering with confidence.
 ---
 
@@ -99,7 +98,8 @@ export default defineTool({
 
 - flat skills can be simple markdown files under `skills/*.md`
 - packaged skills live under `skills/<name>/SKILL.md`
-- packaged `SKILL.md` files must include `name` and `description` frontmatter
+- packaged `SKILL.md` files must include `description` frontmatter; skill identity is derived from
+  the directory name
 - flat skills may omit frontmatter
 
 ## Write Good Descriptions
@@ -145,7 +145,6 @@ If markdown is not enough, you can author a skill in TypeScript with `defineSkil
 import { defineSkill } from "experimental-ash/skills";
 
 export default defineSkill({
-  name: "research",
   description: "Research unfamiliar topics before answering with confidence.",
   markdown:
     "When the task is novel or ambiguous, gather evidence first, then answer with the key facts and the remaining uncertainty.",

package/dist/docs/public/subagents.md

@@ -106,7 +106,7 @@ Child `session.started.data.invocation` includes the parent call/session/turn li
 
 Local subagents may have their own:
 
-- `system`
+- `instructions.md` or `instructions.ts`
 - `skills/`
 - `lib/`
 - `tools/`

package/dist/docs/public/tools.md

@@ -263,7 +263,6 @@ unbounded text, cap it in the executor before returning:
 
 ```ts
 import { defineTool } from "experimental-ash/tools";
-import { truncateHead } from "experimental-ash/sandbox/truncate-output";
 import { z } from "zod";
 
 export default defineTool({

package/dist/docs/public/typescript-api.md

@@ -20,8 +20,9 @@ Source of truth:
 Use these to author the filesystem-backed surface. Each concern has its own subpath:
 `experimental-ash/tools` for tools, `experimental-ash/hooks` for hooks,
 `experimental-ash/sandbox` for the sandbox, `experimental-ash/skills` for skills,
-`experimental-ash/context` for context helpers, and the main `experimental-ash`
-barrel for agent config, schedules, and systems.
+`experimental-ash/schedules` for schedules, `experimental-ash/instructions` for the instructions
+prompt, `experimental-ash/context` for context helpers, and the main `experimental-ash` barrel
+for `defineAgent`.
 
 - `defineAgent(...)` - additive runtime config for `agent.ts`. Subagents reuse the same helper at
   `subagents/<id>/agent.ts`; the only difference is that subagents must declare a `description` so
@@ -41,7 +42,7 @@ barrel for agent config, schedules, and systems.
 
 Framework defaults reachable for spread/wrap composition:
 
-- `bash`, `readFile`, `writeFile`, `webFetch`, `todo`, `loadSkill` (`experimental-ash/tools/defaults`)
+- `bash`, `glob`, `grep`, `readFile`, `writeFile`, `webFetch`, `webSearch`, `todo`, `loadSkill` (`experimental-ash/tools/defaults`)
 
 Most apps use `defineAgent`, `defineTool`, and `defineSandbox` the most.
 
@@ -75,7 +76,7 @@ Channel and Slack types exported from `experimental-ash/channels/slack`:
 - `SlackHandle` - Slack identity handle with `channelId`, `threadTs`, `teamId`, `request()`,
   `uploadFiles()`
 - `SlackThread` - thread-scoped operations: `post`, `postEphemeral`, `startTyping`, `refresh`,
-  `recentMessages`, `mentionUser`, `threadId`
+  `recentMessages`, `mentionUser`
 - `SlackMessage` - parsed inbound mention or DM payload (`text`, `markdown`, `author`,
   `attachments`, `threadTs`, `channelId`, ...)
 - `SlackInteractionAction` - action type for `onInteraction`

package/dist/src/execution/workflow-entry.js

@@ -461,8 +461,8 @@ async function disposeHook(hook) {
  *
  * Used when no explicit continuation token was provided on RunInput
  * (schedule-initiated sessions). The token is derived from the
- * adapter's serialized thread identity — for Slack, this is the
- * thread id `slack:{channelId}:{threadTs}` when threadTs is populated.
+ * adapter's serialized resume identity — for Slack, this is the
+ * continuation token `slack:{channelId}:{threadTs}` when threadTs is populated.
  */
 function deriveAdapterContinuationToken(serializedContext) {
     const channel = serializedContext["ash.channel"];

package/dist/src/harness/prompt-cache.d.ts

@@ -47,12 +47,6 @@ export declare function getAnthropicCacheMarker(): AnthropicCacheMarker;
  * another value.
  */
 export declare function mergeGatewayAutoCaching(base: Readonly<Record<string, unknown>> | undefined): Record<string, unknown>;
-/**
- * Returns a new `providerOptions` object with `gateway.tags` merged into the
- * existing `gateway` sub-object. Preserves authored tags and de-dupes repeated
- * framework tags across retry and compaction paths.
- */
-export declare function mergeGatewayTags(base: Readonly<Record<string, unknown>> | undefined, tags: readonly string[]): Record<string, unknown>;
 /**
  * Returns a new ToolSet where the last tool entry carries the Anthropic
  * cache marker on `providerOptions`. Used on the `anthropic-direct` path

package/dist/src/harness/prompt-cache.js

@@ -1,4 +1,3 @@
-import { createLogger } from "#internal/logging.js";
 /**
  * Shared frozen marker. All direct-Anthropic breakpoints in the harness share
  * this instance to avoid allocating per-message.
@@ -53,75 +52,6 @@ export function mergeGatewayAutoCaching(base) {
         gateway: mergedGateway,
     };
 }
-const MAX_GATEWAY_TAGS = 10;
-const MAX_GATEWAY_TAG_LENGTH = 64;
-const log = createLogger("harness.prompt-cache");
-function normalizeGatewayTag(tag) {
-    const trimmed = tag.trim();
-    if (!trimmed) {
-        return undefined;
-    }
-    return trimmed
-        .replace(/[^a-zA-Z0-9:_./-]+/g, "_")
-        .replace(/_+/g, "_")
-        .slice(0, MAX_GATEWAY_TAG_LENGTH);
-}
-function gatewayTagKey(tag) {
-    const separatorIndex = tag.indexOf(":");
-    return separatorIndex === -1 ? tag : tag.slice(0, separatorIndex);
-}
-/**
- * Returns a new `providerOptions` object with `gateway.tags` merged into the
- * existing `gateway` sub-object. Preserves authored tags and de-dupes repeated
- * framework tags across retry and compaction paths.
- */
-export function mergeGatewayTags(base, tags) {
-    if (tags.length === 0) {
-        return { ...base };
-    }
-    const baseGateway = base?.gateway !== undefined && typeof base.gateway === "object" && base.gateway !== null
-        ? base.gateway
-        : undefined;
-    const authoredTags = Array.isArray(baseGateway?.tags)
-        ? baseGateway.tags.filter((tag) => typeof tag === "string")
-        : [];
-    const mergedTags = [];
-    for (const tag of authoredTags) {
-        const normalized = normalizeGatewayTag(tag);
-        if (!normalized || mergedTags.includes(normalized)) {
-            continue;
-        }
-        mergedTags.push(normalized);
-        if (mergedTags.length === MAX_GATEWAY_TAGS) {
-            break;
-        }
-    }
-    const droppedTags = [];
-    for (const tag of tags) {
-        const normalized = normalizeGatewayTag(tag);
-        if (!normalized || mergedTags.includes(normalized)) {
-            continue;
-        }
-        if (mergedTags.length === MAX_GATEWAY_TAGS) {
-            droppedTags.push(normalized);
-            continue;
-        }
-        mergedTags.push(normalized);
-    }
-    if (droppedTags.length > 0) {
-        log.warn("dropped Ash AI Gateway reporting tags because Gateway tag limit was reached", {
-            droppedTagKeys: [...new Set(droppedTags.map(gatewayTagKey))],
-            limit: MAX_GATEWAY_TAGS,
-        });
-    }
-    return {
-        ...base,
-        gateway: {
-            ...baseGateway,
-            tags: mergedTags,
-        },
-    };
-}
 /**
  * Returns a new ToolSet where the last tool entry carries the Anthropic
  * cache marker on `providerOptions`. Used on the `anthropic-direct` path

package/dist/src/harness/step-hooks.d.ts

@@ -16,7 +16,6 @@ export interface StepHooksInput {
     readonly cachePath: PromptCachePath;
     readonly emit?: HarnessEmitFn;
     readonly emissionState: HarnessEmissionState;
-    readonly gatewayTags?: readonly string[];
     readonly marker: AnthropicCacheMarker | undefined;
     readonly session: HarnessSession;
 }

package/dist/src/harness/step-hooks.js

@@ -2,7 +2,7 @@ import { createActionResultEvent, createActionsRequestedEvent, createStepComplet
 import { createRuntimeToolResultFromMessagePart, createRuntimeToolResultFromStepResult, } from "#harness/action-result-helpers.js";
 import { emitStepStarted, normalizeAssistantStepFinishReason } from "#harness/emission.js";
 import { extractToolApprovalInputRequests } from "#harness/input-extraction.js";
-import { applyConversationCacheControl, mergeGatewayAutoCaching, mergeGatewayTags, } from "#harness/prompt-cache.js";
+import { applyConversationCacheControl, mergeGatewayAutoCaching, } from "#harness/prompt-cache.js";
 import { createRuntimeActionRequestFromToolCall } from "#harness/runtime-actions.js";
 // ---------------------------------------------------------------------------
 // Builder
@@ -41,9 +41,7 @@ export function buildStepHooks(input) {
             messages: processed,
         };
         if (input.cachePath.kind === "gateway-auto") {
-            let providerOptions = mergeGatewayAutoCaching(session.agent.modelReference.providerOptions);
-            providerOptions = mergeGatewayTags(providerOptions, input.gatewayTags ?? []);
-            stepResult.providerOptions = providerOptions;
+            stepResult.providerOptions = mergeGatewayAutoCaching(session.agent.modelReference.providerOptions);
         }
         return stepResult;
     };

package/dist/src/harness/tool-loop.js

@@ -18,7 +18,7 @@ import { getInstrumentationConfig } from "#harness/instrumentation-config.js";
 import { resolveAssistantStepText } from "#harness/messages.js";
 import { classifyModelCallError, summarizeKnownModelCallConfigError, } from "#harness/model-call-error.js";
 import { ensureOtelIntegration } from "#harness/otel-integration.js";
-import { applyLastToolCacheBreakpoint, detectPromptCachePath, getAnthropicCacheMarker, mergeGatewayTags, } from "#harness/prompt-cache.js";
+import { applyLastToolCacheBreakpoint, detectPromptCachePath, getAnthropicCacheMarker, } from "#harness/prompt-cache.js";
 import { createRuntimeActionRequestFromToolCall, resolvePendingRuntimeActions, setPendingRuntimeActionBatch, } from "#harness/runtime-actions.js";
 import { buildStepHooks, emitStepActions, isInvalidToolCall, } from "#harness/step-hooks.js";
 import { pruneToolResults } from "#harness/tool-result-pruning.js";
@@ -105,20 +105,6 @@ function buildGatewayAttributionHeaders(model, runtimeIdentity) {
         headers["http-referer"] = referer;
     return headers;
 }
-function buildGatewayReportingTags({ runtimeIdentity, session, turnId, }) {
-    const tags = ["framework:ash"];
-    if (session.sessionId.startsWith("wrun_")) {
-        tags.push(`workflowRunId:${session.sessionId}`);
-    }
-    tags.push(`sessionId:${session.sessionId}`);
-    if (turnId) {
-        tags.push(`turnId:${turnId}`);
-    }
-    if (runtimeIdentity?.agentId) {
-        tags.push(`agent:${runtimeIdentity.agentId}`);
-    }
-    return tags;
-}
 // ---------------------------------------------------------------------------
 // Turn trace state — survives step boundaries via session.state
 // ---------------------------------------------------------------------------
@@ -262,15 +248,9 @@ export function createToolLoopHarness(config) {
         // Runs before `agent.stream()` so the compacted messages flow through
         // `messages` (which the harness uses to rebuild session history).
         const attributionHeaders = buildGatewayAttributionHeaders(model, config.runtimeIdentity);
-        const gatewayTags = buildGatewayReportingTags({
-            runtimeIdentity: config.runtimeIdentity,
-            session,
-            turnId: emissionState.turnId,
-        });
         ({ messages, session } = await maybeCompact({
             emit,
             emissionState,
-            gatewayTags,
             headers: attributionHeaders,
             messages,
             model,
@@ -304,7 +284,6 @@ export function createToolLoopHarness(config) {
             cachePath,
             emit,
             emissionState,
-            gatewayTags,
             marker,
             session,
         });
@@ -627,13 +606,7 @@ async function maybeCompact(input) {
             usageInputTokens: getInputTokenCount(messages, session.compaction),
         }));
     }
-    const compactionProviderOptions = typeof compaction.model === "string"
-        ? mergeGatewayTags(compaction.providerOptions, [
-            "operation:compaction",
-            ...(input.gatewayTags ?? []),
-        ])
-        : compaction.providerOptions;
-    messages = await compactMessages(messages, compaction.model, session.compaction, compactionProviderOptions, input.telemetry, input.headers);
+    messages = await compactMessages(messages, compaction.model, session.compaction, compaction.providerOptions, input.telemetry, input.headers);
     if (input.onCompaction) {
         const compacted = await input.onCompaction(session);
         session = compacted.session;

package/dist/src/internal/application/package.js

@@ -6,7 +6,7 @@ import { ASH_PACKAGE_NAME } from "#package-name.js";
 let cachedPackageInfo;
 // The package build stamps the published version into `dist` so bundled
 // deployments can still report package metadata without resolving package.json.
-const BUNDLED_FALLBACK_PACKAGE_VERSION = "0.15.0";
+const BUNDLED_FALLBACK_PACKAGE_VERSION = "0.16.1";
 const BUNDLED_FALLBACK_PACKAGE_VERSION_PLACEHOLDER = "__ASH_PACKAGE_VERSION_PLACEHOLDER__";
 const WORKFLOW_MODULE_ALIASES = {
     "workflow/api": "src/compiled/@workflow/core/runtime.js",

package/dist/src/public/channels/slack/api.d.ts

@@ -22,22 +22,12 @@ import { type CardElement, type FileUpload } from "#compiled/chat/index.js";
  */
 export type SlackBotToken = string | (() => string | Promise<string>);
 /**
- * Decodes a Slack thread id of the form `slack:<channelId>:<threadTs>`
- * back into its parts. Returns empty-string fields when the input does
- * not match — used by the channel's fallback paths where the channel
- * id has not yet been observed.
+ * Builds the Slack channel-local continuation token (`<channelId>:<threadTs>`).
+ * Route `send()` namespaces this with the channel name before handing it to the
+ * runtime (`slack:<channelId>:<threadTs>`), so Slack routes should pass the raw
+ * channel-local form here.
  */
-export declare function decodeThreadId(id: string): {
-    channelId: string;
-    threadTs: string;
-};
-/**
- * Builds the Ash thread id (`slack:<channelId>:<threadTs>`) used as the
- * channel's continuation token. Keeps the encoding in one place so the
- * inbound route, the interaction handler, and the context rebuild path
- * cannot drift on the format.
- */
-export declare function encodeThreadId(channelId: string, threadTs: string): string;
+export declare function encodeSlackContinuationToken(channelId: string, threadTs: string): string;
 /**
  * Materializes a {@link SlackBotToken} to a concrete string. Falls
  * back to `process.env.SLACK_BOT_TOKEN` when no value was passed.
@@ -176,8 +166,6 @@ export interface SlackThreadMessage {
  * use {@link SlackHandle.request} on `ctx.slack` instead.
  */
 export interface SlackThread {
-    /** Ash continuation-token thread id — `slack:<channelId>:<threadTs>`. */
-    readonly threadId: string;
     /** Recently fetched thread messages. Populated by {@link refresh}. */
     readonly recentMessages: readonly SlackThreadMessage[];
     /**

package/dist/src/public/channels/slack/api.js

@@ -20,23 +20,13 @@ import { encodeSlackApiBody } from "#public/channels/slack/api-encoding.js";
 import { cardToBlocks, cardToFallbackText } from "#public/channels/slack/blocks.js";
 const log = createLogger("slack.api");
 /**
- * Decodes a Slack thread id of the form `slack:<channelId>:<threadTs>`
- * back into its parts. Returns empty-string fields when the input does
- * not match — used by the channel's fallback paths where the channel
- * id has not yet been observed.
+ * Builds the Slack channel-local continuation token (`<channelId>:<threadTs>`).
+ * Route `send()` namespaces this with the channel name before handing it to the
+ * runtime (`slack:<channelId>:<threadTs>`), so Slack routes should pass the raw
+ * channel-local form here.
  */
-export function decodeThreadId(id) {
-    const parts = id.replace(/^slack:/u, "").split(":");
-    return { channelId: parts[0] ?? "", threadTs: parts[1] ?? "" };
-}
-/**
- * Builds the Ash thread id (`slack:<channelId>:<threadTs>`) used as the
- * channel's continuation token. Keeps the encoding in one place so the
- * inbound route, the interaction handler, and the context rebuild path
- * cannot drift on the format.
- */
-export function encodeThreadId(channelId, threadTs) {
-    return `slack:${channelId}:${threadTs}`;
+export function encodeSlackContinuationToken(channelId, threadTs) {
+    return `${channelId}:${threadTs}`;
 }
 /**
  * Materializes a {@link SlackBotToken} to a concrete string. Falls
@@ -142,7 +132,6 @@ export function buildSlackBinding(input) {
         return { fileIds, raw: complete };
     }
     const thread = {
-        threadId: encodeThreadId(input.channelId, input.threadTs),
         recentMessages: messages,
         async post(rawMessage) {
             const message = normalizePostInput(rawMessage);

package/dist/src/public/channels/slack/interactions.d.ts

@@ -27,11 +27,6 @@ interface ParsedBlockActionsPayload {
     readonly channelId: string;
     readonly threadTs: string;
     readonly teamId: string | undefined;
-    /**
-     * Slack actor that authored the click. Used to attribute the answered
-     * card via `Answered by <@userId>` after a HITL response is delivered.
-     */
-    readonly userId: string | undefined;
     /**
      * The full block list off the clicked message. Preserved on the
      * answered-card update so the original prompt stays visible after the

package/dist/src/public/channels/slack/interactions.js

@@ -17,7 +17,7 @@
  * work runs under `waitUntil` so the webhook ACK is immediate.
  */
 import { createLogger } from "#internal/logging.js";
-import { buildSlackBinding, encodeThreadId, resolveSlackBotToken, } from "#public/channels/slack/api.js";
+import { buildSlackBinding, encodeSlackContinuationToken, resolveSlackBotToken, } from "#public/channels/slack/api.js";
 import { buildAnsweredBlocks, buildFreeformModalView, deriveHitlResponse, freeformRequestIdFromActionId, HITL_FREEFORM_MODAL_ACTION_ID, HITL_FREEFORM_MODAL_BLOCK_ID, HITL_FREEFORM_MODAL_CALLBACK_ID, isFreeformAction, isHitlAction, } from "#public/channels/slack/hitl.js";
 const log = createLogger("slack.interactions");
 /**
@@ -29,30 +29,37 @@ export function parseBlockActionsPayload(body) {
     const actions = body.actions;
     if (!Array.isArray(actions))
         return null;
+    // `channel` and `message` are Optional on block_actions payloads — only
+    // present when the action was triggered from a message in a channel.
     const channel = body.channel?.id;
     const message = body.message;
     const threadTs = message?.thread_ts ?? message?.ts;
     if (!channel || !threadTs)
         return null;
+    // `team` is Required but can be `null` for org-installed apps.
+    // `user` is Required and always carries `id`.
     const team = body.team;
-    const user = body.user;
-    const teamId = team?.id ?? user?.team_id;
-    const userId = typeof user?.id === "string" ? user.id : undefined;
-    const messageTs = typeof message?.ts === "string" ? message.ts : undefined;
-    const messageBlocks = Array.isArray(message?.blocks) ? message.blocks : [];
+    const userBlock = body.user;
+    const teamId = team?.id ?? userBlock.team_id;
+    const user = {
+        id: userBlock.id,
+        username: userBlock.username,
+        name: userBlock.name,
+    };
+    const messageBlocks = message?.blocks ?? [];
     return {
         actions: actions.map((a) => ({
             actionId: String(a.action_id ?? ""),
             value: a.value != null ? String(a.value) : undefined,
             blockId: a.block_id != null ? String(a.block_id) : undefined,
             selectedOptionValue: extractSelectedOptionValue(a),
-            messageTs,
+            messageTs: message?.ts,
             label: extractActionLabel(a),
+            user,
         })),
         channelId: channel,
         threadTs,
         teamId,
-        userId,
         messageBlocks,
     };
 }
@@ -117,7 +124,7 @@ export async function handleInteractionPost(rawBody, ctx, deps) {
         await openFreeformModal({ payload, interaction, freeformAction, deps });
         return ack;
     }
-    const continuationToken = encodeThreadId(interaction.channelId, interaction.threadTs);
+    const continuationToken = encodeSlackContinuationToken(interaction.channelId, interaction.threadTs);
     const inputResponses = interaction.actions
         .map(deriveHitlResponse)
         .filter((r) => r !== null);
@@ -130,7 +137,7 @@ export async function handleInteractionPost(rawBody, ctx, deps) {
                 channelId: interaction.channelId,
                 threadTs: interaction.threadTs,
                 teamId: interaction.teamId ?? null,
-                triggeringUserId: interaction.userId ?? null,
+                triggeringUserId: interaction.actions[0]?.user.id ?? null,
             },
         })
             .catch((error) => {
@@ -177,7 +184,7 @@ async function openFreeformModal(input) {
         return;
     }
     const metadata = {
-        continuationToken: encodeThreadId(input.interaction.channelId, input.interaction.threadTs),
+        continuationToken: encodeSlackContinuationToken(input.interaction.channelId, input.interaction.threadTs),
         channelId: input.interaction.channelId,
         threadTs: input.interaction.threadTs,
         messageTs,
@@ -222,9 +229,11 @@ async function handleViewSubmission(payload, ctx, _deps) {
     const text = typeof raw === "string" ? raw : "";
     if (text.length === 0)
         return ack;
+    // `user` is Required on view_submission payloads; `team_id` is on the
+    // user object in modern payloads but not guaranteed in all examples.
     const user = payload.user;
-    const triggeringUserId = typeof user?.id === "string" ? user.id : null;
-    const teamId = typeof user?.team_id === "string" ? user.team_id : null;
+    const triggeringUserId = user.id;
+    const teamId = user.team_id ?? null;
     ctx.waitUntil(ctx
         .send({ inputResponses: [{ requestId: metadata.requestId, text }] }, {
         auth: null,
@@ -260,7 +269,7 @@ async function updateAnsweredHitlCard(interaction, deps) {
     const blocks = buildAnsweredBlocks({
         promptBlock: findPromptBlock(interaction.messageBlocks),
         answerLabel,
-        userId: interaction.userId,
+        userId: hitlAction.user.id,
     });
     const token = await resolveSlackBotToken(deps.config.credentials?.botToken);
     const response = await fetch("https://slack.com/api/chat.update", {

package/dist/src/public/channels/slack/slackChannel.d.ts

@@ -18,7 +18,7 @@ type EventData<T extends HandleMessageStreamEvent["type"]> = Extract<HandleMessa
  *
  * - {@link thread} owns thread-scoped operations (`post`,
  *   `postEphemeral`, `startTyping`, `refresh`, `recentMessages`,
- *   `threadId`, `mentionUser`). Reads as "post a reply to this thread".
+ *   `mentionUser`). Reads as "post a reply to this thread".
  * - {@link slack} owns Slack identity (`channelId`, `threadTs`,
  *   `teamId`) plus the raw-API escape hatch (`request`, `uploadFiles`).
  */
@@ -117,6 +117,24 @@ export interface SlackInteractionAction {
      * the "answered" card without re-fetching the original request.
      */
     readonly label?: string;
+    /**
+     * Slack actor who triggered the interaction. Lets `onInteraction`
+     * handlers attribute resolutions back to the clicker (e.g. "Filed by
+     * <@U0123456789>") without re-parsing the raw payload. Always present
+     * — Slack requires `user` on every `block_actions` payload.
+     */
+    readonly user: SlackInteractionUser;
+}
+/**
+ * Slack actor surfaced on {@link SlackInteractionAction.user}. Mirrors
+ * `body.user` from the inbound `block_actions` payload.
+ */
+export interface SlackInteractionUser {
+    readonly id: string;
+    /** Modern canonical display handle. */
+    readonly username?: string;
+    /** Legacy display handle, kept for older workspaces. */
+    readonly name?: string;
 }
 /**
  * Result of an `onAppMention` or `onDirectMessage` callback. Return

package/dist/src/public/channels/slack/slackChannel.js

@@ -1,5 +1,5 @@
 import { createLogger } from "#internal/logging.js";
-import { buildSlackBinding, encodeThreadId, } from "#public/channels/slack/api.js";
+import { buildSlackBinding, encodeSlackContinuationToken, } from "#public/channels/slack/api.js";
 import { buildSlackTurnMessage, collectInboundFileParts, createSlackFetchFile, } from "#public/channels/slack/attachments.js";
 import { defaultEvents, defaultInputRequestedHandler, defaultOnAppMention, defaultOnDirectMessage, } from "#public/channels/slack/defaults.js";
 import { parseAppMentionEvent, parseDirectMessageEvent, prependSlackContext, } from "#public/channels/slack/inbound.js";
@@ -79,7 +79,7 @@ export function slackChannel(config = {}) {
             const threadTs = typeof input.args.threadTs === "string" ? input.args.threadTs : "";
             return send(input.message, {
                 auth: input.auth,
-                continuationToken: encodeThreadId(channelId, threadTs),
+                continuationToken: encodeSlackContinuationToken(channelId, threadTs),
                 state: {
                     channelId,
                     threadTs: threadTs || null,
@@ -184,7 +184,7 @@ async function dispatchInboundMessage(input) {
     }
     if (result === null || result === undefined)
         return;
-    const continuationToken = encodeThreadId(message.channelId, message.threadTs);
+    const continuationToken = encodeSlackContinuationToken(message.channelId, message.threadTs);
     const fileParts = await collectInboundFileParts({
         mention: message,
         thread,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "experimental-ash",
-  "version": "0.15.0",
+  "version": "0.16.1",
   "bin": {
     "ash": "./bin/ash.js",
     "experimental-ash": "./bin/ash.js"