sparkecoder 0.1.123 → 0.1.125

package/dist/skills/default/memory.md

@@ -0,0 +1,49 @@
+---
+name: Memory
+description: Read, update, organize, or explain durable SparkECoder memory. Load when the user asks the agent to remember something, create memory files, edit its memory, or retrieve long-term context.
+---
+
+# Memory
+
+SparkECoder memory is filesystem-backed markdown. Agents can read and edit it with normal file tools.
+
+## Default Layout
+
+- `.sparkecoder/rules/memory.md` - always-loaded memory index. Keep it short.
+- `.sparkecoder/memory/` - detailed durable memory files.
+
+The memory index can point to any deeper file. Create folders under `.sparkecoder/memory/` when a topic grows.
+
+Suggested structure:
+
+```text
+.sparkecoder/memory/
+  preferences.md
+  project.md
+  operations.md
+  decisions/
+  integrations/
+  runbooks/
+```
+
+## Rules
+
+- Do not store secrets, credentials, API keys, private tokens, or sensitive personal data.
+- Store durable facts, preferences, architecture decisions, runbooks, and pointers.
+- Keep `.sparkecoder/rules/memory.md` as low-token as possible. It may have `contextBudgetTokens` frontmatter; if content exceeds that budget, SparkECoder truncates it in the system prompt. A large budget is a ceiling, not a target.
+- Put long content in `.sparkecoder/memory/**/*.md` and leave a short pointer in the index.
+- Read existing memory before editing it.
+- Remove or update stale memory instead of appending contradictions.
+
+## Remember Something
+
+1. Decide whether it belongs in the always-loaded index or a detailed memory file.
+2. If it is short and should always be known, update `.sparkecoder/rules/memory.md`.
+3. If it is detailed, create or edit a file under `.sparkecoder/memory/` and add only a pointer to the index.
+4. Keep wording factual and timestamp-free unless the date is part of the durable fact.
+
+## Retrieve Memory
+
+1. Read `.sparkecoder/rules/memory.md`.
+2. Follow the relevant pointer into `.sparkecoder/memory/`.
+3. Search within `.sparkecoder/memory/` if the index points to a broad area.

package/dist/skills/default/recording.md

@@ -51,7 +51,7 @@ sparkecoder record stop-all       # stop everything (cleanup)
 sparkecoder record start --dir /custom/path --name my-task
 ```
 
-Storage: defaults to `~/recordings/`. Files are named `rec-<id>[-<name>].<mov|mp4>`.
+Storage: defaults to `~/recordings/`. Files are named `rec-<id>[-<name>].mp4`. On macOS the helper captures a temporary `.mov` and converts it to `.mp4` when stopped.
 
 ### Why not `screencapture -V N` or raw `ffmpeg` backgrounded?
 
@@ -190,7 +190,7 @@ For a **task worker** that captured a recording, include the structured fields b
 {
   "summary": "Opened Calculator, typed 2+2, hit Return. Got 4.",
   "recording": {
-    "path": "/Users/me/recordings/rec-abc123-calculator-demo.mov",
+    "path": "/Users/me/recordings/rec-abc123-calculator-demo.mp4",
     "durationSec": 42,
     "sizeMb": 18.4,
     "description": "Full screen recording of the Calculator interaction"

package/dist/skills/default/skill-authoring.md

@@ -0,0 +1,96 @@
+---
+name: Skill Authoring
+description: Create, edit, debug, or explain SparkECoder skills and rules. Load when the user asks to add a skill, edit a skill, teach agents a workflow, update skill discovery, or explain where skill files live.
+---
+
+# Skill Authoring
+
+SparkECoder skills are filesystem-backed markdown files. Use normal file tools (`read_file`, `write_file`, `bash`) to inspect, create, edit, and verify them.
+
+## Skill Locations
+
+Discovery starts from the configured working directory. Higher priority wins when two skills share the same name.
+
+1. `.sparkecoder/rules/` - always loaded into every agent turn.
+2. `.sparkecoder/skills/` - on-demand skills for this project.
+3. `.cursor/rules/` - Cursor-style rule files; frontmatter can make them always loaded.
+4. `.claude/skills/` - Claude/Cursor skill directories with `SKILL.md`.
+5. `skills/` - legacy project skills.
+6. `src/skills/default/` - built-in SparkECoder skills shipped with this package.
+
+Configured `skills.additionalDirectories` in `sparkecoder.config.json` are also loaded after these discovered directories.
+
+## Preferred Placement
+
+- Project-specific behavior: create `.sparkecoder/skills/<skill-name>.md` or `.sparkecoder/skills/<skill-name>/SKILL.md`.
+- Always-on project guidance: create `.sparkecoder/rules/<rule-name>.md`.
+- Built-in behavior for all SparkECoder installs: edit or add `src/skills/default/<skill-name>.md`.
+- Legacy compatibility only: use `skills/`.
+
+Do not create skills in Cursor's internal `~/.cursor/skills-cursor/` directory.
+
+## File Formats
+
+Flat markdown skill:
+
+```markdown
+---
+name: Skill Name
+description: What this skill does and when to load it.
+---
+
+# Skill Name
+
+Instructions...
+```
+
+Directory skill:
+
+```text
+.sparkecoder/skills/skill-name/
+  SKILL.md
+  reference.md
+  examples.md
+```
+
+Keep `SKILL.md` concise. Put long references or examples in sibling files and link to them by relative path.
+
+## Frontmatter
+
+Supported fields:
+
+- `name`: Display/load name. Matching is case-insensitive.
+- `description`: Appears in the available-skills list; include trigger terms.
+- `alwaysApply: true`: Always inject this skill.
+- `globs`: Auto-load when active files match patterns.
+- `platforms`: Optional process platforms (`darwin`, `linux`, `win32`).
+
+Example:
+
+```markdown
+---
+name: React Doctor
+description: Check React code for common bugs after UI edits. Load after changing React components, hooks, or client state.
+globs: ["*.tsx", "*.jsx", "web/src/**"]
+---
+```
+
+## Workflow
+
+1. Inspect existing skills with `ls` or file search before adding a new one.
+2. Prefer editing an existing skill if it already owns the workflow.
+3. Use a clear lowercase filename such as `skill-authoring.md`.
+4. Write a specific description that says both what the skill does and when to load it.
+5. Keep instructions action-oriented and repo-specific.
+6. Verify discovery by running the focused skills tests or loading the skill by name.
+
+## Verification
+
+Useful checks:
+
+```bash
+pnpm vitest run tests/skills.test.ts
+pnpm run typecheck
+```
+
+If you edited built-in skills, rebuild before expecting installed CLIs to see the change.

package/dist/tools/index.d.ts

@@ -1,7 +1,7 @@
 import * as ai from 'ai';
 import { ToolSet } from 'ai';
-import { B as BashToolProgress, W as WriteFileProgress, S as SearchToolProgress } from '../search-DOzC4ojH.js';
-export { b as BashToolOptions, d as SearchToolOptions, e as WriteFileToolOptions, c as createBashTool, f as createSearchTool, a as createWriteFileTool } from '../search-DOzC4ojH.js';
+import { B as BashToolProgress, W as WriteFileProgress, S as SearchToolProgress } from '../search-CVVfuBPZ.js';
+export { b as BashToolOptions, d as SearchToolOptions, e as WriteFileToolOptions, c as createBashTool, f as createSearchTool, a as createWriteFileTool } from '../search-CVVfuBPZ.js';
 
 interface TaskCompletionSignal {
     status: 'completed' | 'failed';
@@ -110,7 +110,7 @@ interface TodoToolOptions {
     workingDirectory: string;
 }
 declare function createTodoTool(options: TodoToolOptions): ai.Tool<{
-    action: "list" | "add" | "mark" | "clear" | "save_plan" | "list_plans" | "get_plan" | "delete_plan";
+    action: "clear" | "add" | "list" | "mark" | "save_plan" | "list_plans" | "get_plan" | "delete_plan";
     status?: "completed" | "pending" | "in_progress" | "cancelled" | undefined;
     items?: {
         content: string;

package/dist/tools/index.js

@@ -32,7 +32,10 @@ var init_types = __esm({
       // not listed here. Values match `process.platform`
       // (darwin, linux, win32, freebsd, ...). If omitted or empty, the skill is
       // available on all platforms.
-      platforms: z.array(z.string()).optional().default([])
+      platforms: z.array(z.string()).optional().default([]),
+      // Optional approximate token budget for always-loaded content. If set,
+      // the prompt builder truncates this skill/rule before injecting it.
+      contextBudgetTokens: z.number().int().positive().optional()
     });
     TaskConfigSchema = z.object({
       enabled: z.boolean(),
@@ -3270,6 +3273,8 @@ function parseSkillFrontmatter(content) {
           data[key] = true;
         } else if (value === "false") {
           data[key] = false;
+        } else if (/^\d+$/.test(value)) {
+          data[key] = Number(value);
         } else {
           data[key] = value;
         }
@@ -3329,7 +3334,8 @@ async function loadSkillsFromDirectory(directory, options = {}) {
         loadType,
         priority,
         sourceDir: directory,
-        platforms: parsed.metadata.platforms
+        platforms: parsed.metadata.platforms,
+        contextBudgetTokens: parsed.metadata.contextBudgetTokens
       });
     } else {
       const name = getSkillNameFromPath(filePath);
@@ -3343,7 +3349,8 @@ async function loadSkillsFromDirectory(directory, options = {}) {
         loadType: forceAlwaysApply ? "always" : defaultLoadType,
         priority,
         sourceDir: directory,
-        platforms: []
+        platforms: [],
+        contextBudgetTokens: void 0
       });
     }
   }
@@ -3385,11 +3392,12 @@ function formatSkillsForContext(skills) {
   if (onDemandSkills.length === 0) {
     return "No on-demand skills available.";
   }
-  const lines = ["Available skills (use load_skill tool to load into context):"];
+  const lines = ["<available_skills>", "Use the load_skill tool to load one of these into context:"];
   for (const skill of onDemandSkills) {
     const globInfo = skill.globs?.length ? ` [auto-loads for: ${skill.globs.join(", ")}]` : "";
     lines.push(`- ${skill.name}: ${skill.description}${globInfo}`);
   }
+  lines.push("</available_skills>");
   return lines.join("\n");
 }