akm-opencode 0.9.25202610050121 → 0.9.27202610072331

package/README.md

@@ -1,6 +1,6 @@
 # akm-opencode
 
-OpenCode plugin for [AKM](https://github.com/itlackey/akm) `0.9.25`. It exposes exactly five public tools and uses lifecycle hooks to bring relevant AKM context into a session. The dependency is exact-pinned because the plugin calls AKM's in-process `dist/` APIs.
+OpenCode plugin for [AKM](https://github.com/itlackey/akm) `0.9.27`. It exposes exactly five public tools and uses lifecycle hooks to bring relevant AKM context into a session. The dependency is exact-pinned because the plugin calls AKM's in-process `dist/` APIs.
 
 ## Installation
 

package/dist/index.js

@@ -2354,11 +2354,11 @@ var AKM_HINTS_PREFIX = [
   "# AKM is available in this session",
   "",
   "- Assets for this project (skills, knowledge, memories, workflows) live in the AKM bundle. Before writing or editing a file whose format or keys you are not sure of, find them with `akm_curate` (by task) or `akm_search` (by name), and read one with `akm_show` before relying on it.",
-  "- Record `akm_feedback` on an asset once you know whether it helped; keep durable project knowledge with `akm_remember`.",
+  "- Record `akm_feedback` on an asset when it helped, or when its content proved wrong or stale; keep durable project knowledge with `akm_remember`.",
   "- `akm hints` has this bundle's conventions; `akm help` has the CLI."
 ].join(`
 `);
-var AKM_CURATED_TAIL = "\n\nTip: call `akm_show <ref>` to fetch full content. If an asset is wrong, stale or incomplete, call `akm_feedback <ref> negative` with a note saying <what is wrong and what should change>: that queues it for review and a fix. Use `positive` when it helped.";
+var AKM_CURATED_TAIL = "\n\nTip: call `akm_show <ref>` to fetch full content. Only if an asset is wrong or stale, call `akm_feedback <ref> negative` with a note saying <what is wrong and what it should say>: that lowers its ranking. To correct a fact you have verified, also pass `replace`, `with` and `source`. Use `positive` when it helped. An asset that simply didn't fit your task is not negative feedback: record nothing.";
 var AKM_CONTEXT_TRUNCATED_MARKER = `
 
 [truncated for context]`;
@@ -3364,7 +3364,7 @@ ${hint}` : hint;
         }
       }),
       akm_show: tool({
-        description: "Show an AKM asset by [bundle//]conceptId[#fragment]. Read an asset this way before relying on it, then record akm_feedback once you know whether it helped.",
+        description: "Show an AKM asset by [bundle//]conceptId[#fragment]. Read an asset this way before relying on it, then record akm_feedback when it helped, or when its content proved wrong or stale.",
         args: {
           ref: tool.schema.string().describe("Asset ref returned by akm_curate or akm_search, optionally with a #fragment \u2014 e.g. `skills/code-review` or `local//knowledge/deploy#Rollback`."),
           detail: tool.schema.enum(["brief", "summary", "normal", "full"]).optional().describe("Response detail level. Defaults to 'normal'.")
@@ -3391,16 +3391,44 @@ ${hint}` : hint;
         }
       }),
       akm_feedback: tool({
-        description: "Record feedback for a bundle asset. Negative feedback with a note queues the asset for review: the next improve run proposes a fix based on the note, so say what is wrong and what should change. Positive feedback raises the asset's ranking and never triggers a rewrite. Call it after akm_show when the asset's content materially helped, or proved wrong, stale or incomplete. A failed akm call is not feedback on the asset.",
+        description: "Record feedback for a bundle asset. Negative feedback is only for content that is wrong or stale. With a note it flags the asset and lowers its ranking; the next improve run may repair only its description, title or when_to_use from the note, so say what is wrong and what it should say. To correct a wrong fact in its text, also pass replace, with and source: akm checks the fix and queues it as a proposal for review. Attach a fix only when you have verified the correct fact (ran the command, read the official doc or the source file), otherwise record the note only. Positive feedback only raises the asset's ranking. An asset that simply didn't fit your task is not negative feedback: record nothing. Call it after akm_show when the asset's content materially helped, or proved wrong or stale. A failed akm call is not feedback on the asset.",
         args: {
           ref: tool.schema.string().describe("Asset ref to record feedback for."),
           sentiment: tool.schema.enum(["positive", "negative"]).describe("Whether the feedback is positive or negative."),
-          note: tool.schema.string().optional().describe("What is wrong and what should change. Required for negative feedback; it guides the fix.")
+          note: tool.schema.string().optional().describe("What is wrong and what it should say. Required for negative feedback."),
+          replace: tool.schema.array(tool.schema.string()).optional().describe("Exact current text to correct, copied verbatim from the asset's file (akm_show returns its path). Each must appear exactly once there. Pair each with a with entry, in order. Negative feedback only."),
+          with: tool.schema.array(tool.schema.string()).optional().describe("The corrected text for each replace entry, in order. Change only the wrong words or lines: no rewording, no added headings or intros."),
+          source: tool.schema.string().optional().describe("The URL, command or file that shows the correct fact. Required with replace.")
         },
-        async execute({ ref, sentiment, note }, context) {
+        async execute({ ref, sentiment, note, replace, with: corrections, source }, context) {
+          const fixes = replace?.length ?? 0;
+          const corrected = corrections?.length ?? 0;
+          let refusal;
+          if (fixes !== corrected)
+            refusal = `Each replace needs one with (got ${fixes} replace and ${corrected} with).`;
+          else if (fixes > 0 && sentiment !== "negative")
+            refusal = "replace, with and source are only for negative feedback.";
+          else if (fixes > 0 && !source?.trim())
+            refusal = "A fix needs source: the URL, command or file that shows the correct fact.";
+          if (refusal) {
+            await writePluginLog(logClient, "warn", "AKM feedback refused", {
+              subsystem: "feedback",
+              toolName: "akm_feedback",
+              sessionID: context.sessionID,
+              directory: context.directory,
+              ref,
+              sentiment,
+              error: refusal
+            });
+            return JSON.stringify({ ok: false, error: refusal });
+          }
           const args = ["feedback", ref, sentiment === "positive" ? "--positive" : "--negative"];
           if (note)
             args.push("--reason", note);
+          for (let i = 0;i < fixes; i++)
+            args.push("--replace", replace[i], `--with=${corrections[i]}`);
+          if (source)
+            args.push("--source", source);
           const raw = await runCli(client, args, { toolName: "akm_feedback", sessionID: context.sessionID, directory: context.directory });
           const parsed = safeJsonParse(raw);
           if (parsed?.ok === false) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "akm-opencode",
-  "version": "0.9.25202610050121",
+  "version": "0.9.27202610072331",
   "type": "module",
   "description": "OpenCode plugin for AKM with discovery tools, feedback, memory, and automatic learning proposals from corrections and recurring workflows.",
   "keywords": [
@@ -38,7 +38,7 @@
   },
   "dependencies": {
     "@opencode-ai/plugin": "^1.18.14",
-    "akm-cli": "0.9.25"
+    "akm-cli": "0.9.27"
   },
   "devDependencies": {
     "@types/node": "^26.1.1",