@oh-my-pi/pi-coding-agent 3.24.0 → 3.25.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (20) hide show
  1. package/CHANGELOG.md +15 -0
  2. package/package.json +4 -4
  3. package/src/core/custom-commands/bundled/wt/index.ts +3 -0
  4. package/src/core/sdk.ts +7 -0
  5. package/src/core/tools/complete.ts +131 -0
  6. package/src/core/tools/index.test.ts +9 -1
  7. package/src/core/tools/index.ts +18 -5
  8. package/src/core/tools/jtd-to-json-schema.ts +274 -0
  9. package/src/core/tools/output.ts +125 -14
  10. package/src/core/tools/task/artifacts.ts +6 -9
  11. package/src/core/tools/task/executor.ts +44 -5
  12. package/src/core/tools/task/index.ts +23 -18
  13. package/src/core/tools/task/name-generator.ts +247 -0
  14. package/src/core/tools/task/render.ts +137 -8
  15. package/src/core/tools/task/types.ts +7 -0
  16. package/src/core/tools/task/worker-protocol.ts +1 -0
  17. package/src/core/tools/task/worker.ts +33 -1
  18. package/src/prompts/task.md +14 -50
  19. package/src/prompts/tools/output.md +2 -1
  20. package/src/prompts/tools/task.md +3 -1
package/src/prompts/tools/task.md CHANGED
@@ -16,8 +16,9 @@ The Task tool launches specialized agents (workers) that autonomously handle com
16
16
  ## Usage Notes
17
17
 
18
18
  - Always include a short description of the task in the task parameter
19
- - **Plan-then-execute**: Put shared constraints in `context`, keep each task focused, specify output format and acceptance criteria
19
+ - **Plan-then-execute**: Put shared constraints in `context`, keep each task focused, specify acceptance criteria; use `output_schema` when you need structured output
20
20
  - **Minimize tool chatter**: Avoid repeating large context; use Output tool with output ids for full logs
21
+ - **Structured completion**: If `output_schema` is provided, subagents must call `complete` to finish
21
22
  - **Parallelize**: Launch multiple agents concurrently whenever possible
22
23
  - **Results are intermediate data**: Agent findings provide context for YOU to perform actual work. Do not treat agent reports as "task complete" signals.
23
24
  - **Stateless invocations**: Each agent runs autonomously and returns a single final message. Include all necessary context and specify exactly what information to return.
@@ -30,6 +31,7 @@ The Task tool launches specialized agents (workers) that autonomously handle com
30
31
  - `tasks`: Array of `{agent, task, description?, model?}` - tasks to run in parallel (max {{MAX_PARALLEL_TASKS}}, {{MAX_CONCURRENCY}} concurrent)
31
32
  - `model`: (optional) Override the agent's default model with fuzzy matching (e.g., "sonnet", "codex", "5.2"). Supports comma-separated fallbacks: "gpt, opus" tries gpt first, then opus. Use "default" for omp's default model
32
33
  - `context`: (optional) Shared context string prepended to all task prompts - use this to avoid repeating instructions
34
+ - `output_schema`: (optional) JSON schema for structured subagent output (used by the complete tool)
33
35
 
34
36
  ## Examples
35
37