@akagilnc/pi-workflow-roles 0.1.2460 → 0.1.2464

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.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@akagilnc/pi-workflow-roles",
3
- "version": "0.1.2460",
3
+ "version": "0.1.2464",
4
4
  "description": "Soul-bound workflow roles for Pi",
5
5
  "type": "module",
6
6
  "license": "Apache-2.0",
@@ -7,26 +7,40 @@ Before invoking the engine, read `../engine-dispatch.md`, resolving that path
7
7
  relative to this note. This note only covers this engine's CLI technical
8
8
  parameters.
9
9
 
10
- ## Invocation examples (local Cursor Agent CLI)
10
+ ## Entrypoint (verified 2026-08-28)
11
11
 
12
- The machine entrypoint is `agent` (also exposed as `cursor-agent`). Run from the
13
- role project root. Non-interactive print mode:
12
+ The machine entrypoint is **`cursor-agent`**. Do NOT use the bare `agent`
13
+ name on this host: PATH resolves `agent` to the Grok CLI (`~/.grok/bin/agent`
14
+ → grok 1.0.5, a different product billed to grok.com Build) — its flags
15
+ differ and it fails labor invocations (`--force` unknown; billing 402 when
16
+ the Grok Build balance is exhausted).
14
17
 
15
- ```bash
16
- agent -p --force --output-format text "YOUR_LABOR_PROMPT"
17
- ```
18
+ ## Invocation (local Cursor Agent CLI)
18
19
 
19
- Choose a model explicitly when the seat needs a known Cursor model id:
20
+ Run from the role project root. Non-interactive print mode; `-f/--force`
21
+ allows command execution headlessly; `-p` prints responses for scripts:
20
22
 
21
23
  ```bash
22
- agent -p --force --model grok-4.5 --output-format text "YOUR_LABOR_PROMPT"
24
+ cursor-agent -p -f --output-format text --model <MODEL_ID> "YOUR_LABOR_PROMPT"
23
25
  ```
24
26
 
25
- Stream JSON events for long labor:
27
+ - Model ids come from `cursor-agent models`. Effort tiers are encoded in the
28
+ id itself (e.g. `cursor-grok-4.6-low`, `cursor-grok-4.6-medium`,
29
+ `cursor-grok-4.6-high`, plus `-fast` variants); some models also accept a
30
+ bracket override form (`'claude-opus-4-8[context=1m,effort=high]'` — see
31
+ `cursor-agent --help`).
32
+ - Owner pool directive 2026-08-28: default labor model = `cursor-grok-4.6-low`.
33
+ - Stream JSON events for long labor: `--output-format stream-json`.
34
+
35
+ Prefer `cursor-agent --help` on the host over any remembered flag set. Do not
36
+ wrap this engine behind `ak-role` flags.
37
+
38
+ ## Smoke test (run before first labor leg of a session)
26
39
 
27
40
  ```bash
28
- agent -p --force --output-format stream-json "YOUR_LABOR_PROMPT"
41
+ cursor-agent -p -f --output-format text --model cursor-grok-4.6-low "Reply with exactly one word: OK"
29
42
  ```
30
43
 
31
- Prefer `agent --help` on the host over any remembered flag set. Do not wrap this
32
- engine behind `ak-role` flags.
44
+ Expected: stdout ends with exactly `OK`, exit code 0. Verified 2026-08-28 on
45
+ this host (Cursor subscription login). If it asks about directory trust, the
46
+ `-f` flag (or `--trust`) is missing.
@@ -65,9 +65,8 @@ hermes -z "YOUR_LABOR_PROMPT" --in /path/to/project --no-restore-cwd \
65
65
  simultaneously, losing legs saw sustained HTTP 429 across 9-29 detour
66
66
  attempts while the 1-2 winners completed normally (Ming_LLM books runs
67
67
  01a046ec-*/01a046f4-*, 14:55-15:04 window).
68
- - Keep at most ONE in-flight hermes labor call per account; queue the rest.
69
- - On 429: back off once (~60s) and retry once; if still 429, stop and record
70
- the typed engine failure in the receipt — do not hammer retries.
68
+ No concurrency ceiling has been measured beyond this observation; policy on
69
+ concurrency and retries is the dispatcher's call, not this note's.
71
70
 
72
71
  ## Smoke test (run before first labor leg of a session)
73
72