You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit d55008a
Browse filesBrowse the repository at this point in the historyBrowse files
Copy file name to clipboardExpand all lines: .agents/skills/benchmark-coder/SKILL.md
+15-3Lines changed: 15 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -21,12 +21,24 @@ Use the runtime helpers explicitly in authored benchmark scripts:
21
21
22
22
```python
23
23
from utils.submission import validate, simulate, submit_for_review
24
-
from utils.preview import preview, preview_drawing, objectives_geometry
24
+
from utils.preview import (
25
+
list_render_bundles,
26
+
objectives_geometry,
27
+
pick_preview_pixel,
28
+
pick_preview_pixels,
29
+
preview,
30
+
preview_drawing,
31
+
query_render_bundle,
32
+
)
25
33
```
26
34
27
35
-`validate(result)` and `simulate(result)` are the required pre-handoff checks.
28
36
-`submit_for_review(result)` is the final benchmark review handoff helper.
29
-
-`preview(...)` and `preview_drawing()` are the evidence-generation paths; `objectives_geometry()` reconstructs benchmark objective overlays when needed.
37
+
-`preview(...)` is the live scene and objective-overlay path; use `payload_path=True` only when the current workflow needs the live payload-path overlay. `preview_drawing()` is the drafting-package path and keeps the payload overlay off.
38
+
-`objectives_geometry()` reconstructs benchmark objective overlays when needed.
39
+
-`list_render_bundles()` selects the exact current or historical render bundle before you inspect media or point-pick results.
40
+
-`query_render_bundle()` returns compact bundle metadata and frame slices without pulling the full media payload.
41
+
-`pick_preview_pixel()` and `pick_preview_pixels()` resolve screen-space points against the bundle-local snapshot when click-to-world evidence matters.
30
42
-`from utils.visualize import ...` is a compatibility alias, but `utils.preview` is the preferred namespace for new code.
31
43
32
44
## Geometry Contract
@@ -105,7 +117,7 @@ Do not invent fallback behavior to paper over contradictions. If the approved pl
105
117
4. Run `python benchmark_script.py` as the canonical execution path.
106
118
5. Fix import, geometry, labeling, or contract failures first.
107
119
6. Run validation and simulation on the latest revision.
108
-
7. Inspect render images or simulation media when they exist or when motion behavior is uncertain.
120
+
7. Inspect render images or simulation media when they exist or when motion behavior is uncertain. If the question depends on bundle identity or pixel-to-world mapping, select the exact bundle with `list_render_bundles()` and query that bundle-local snapshot with `query_render_bundle()` or `pick_preview_pixel()` before changing geometry.
109
121
8. Keep `todo.md` and `journal.md` synchronized with the work actually being done.
110
122
9. Submit for review only when the latest revision is valid, simulated, and ready.
-[ ] Strong reject plans that cannot state the payload trajectory clearly or that fail to derive it rigorously in `plan.md`; a missing trajectory estimate usually means downstream machinery is not grounded either.
38
41
-[ ] Confirm `plan.md`'s detailed calculation section fully explains the numbers used to justify the precise path, including intermediate derivations for distances, timing, speeds, and clearances.
39
42
-[ ] If render or drawing evidence exists, inspect it with `inspect_media(...)`. If drafting evidence must be materialized first, call `preview_drawing()` and inspect the persisted output.
43
+
-[ ] If bundle identity or a pixel-to-world question matters, resolve the exact bundle with `list_render_bundles()` and query that bundle-local snapshot before deciding.
40
44
-[ ] Treat `benchmark_plan_evidence_script.py` and `benchmark_plan_technical_drawing_script.py` as the inspectable source of the approved benchmark solution, not just helper files.
41
45
-[ ] Verify cross-artifact consistency for labels, repeated quantities, COTS identities, zone geometry, randomization, bounds, runtime jitter, and motion facts. See `references/review_contracts.md`.
42
46
-[ ] Verify `moved_object.material_id` resolves to a known material and `benchmark_assembly_definition.yaml` is a schema-valid full `AssemblyDefinition`.
short_description: "Review benchmark planner handoffs against strict geometry, motion, precise-path, and scientifically derived calculation contracts."
4
-
default_prompt: "Use this skill when reviewing benchmark planner handoffs before coding starts. Read the latest planner artifacts, the stage manifest, and the relevant architecture docs; follow the checklist in `SKILL.md` and `references/review_contracts.md`; inspect any render or drawing evidence with inspect_media(...) and preview_drawing() when needed; compare `assembly_definition.yaml.motion_forecast` to `payload_trajectory_definition.yaml` when present and decide whether the declared path is physically satisfiable; treat a missing or hand-wavy payload trajectory estimate as a strong reject because downstream machinery timing and sizing are then ungrounded; require `plan.md`'s detailed calculation section to fully derive the path-relevant distances, times, speeds, clearances, and tolerances; treat the planner evidence and technical-drawing scripts as the source of the approved benchmark contract; verify exact label, quantity, COTS identity, motion, timing, speed plausibility, and schema grounding; and write the stage-scoped decision/comments YAML pair under reviews/ before submitting the review gate."
4
+
default_prompt: "Use this skill when reviewing benchmark planner handoffs before coding starts. Read the latest planner artifacts, the stage manifest, and the relevant architecture docs; follow the checklist in `SKILL.md` and `references/review_contracts.md`; inspect any render or drawing evidence with inspect_media(...) and preview_drawing() when needed, and use list_render_bundles(), query_render_bundle(), or pick_preview_pixel() when the review depends on exact bundle history or a screen-space point; compare `assembly_definition.yaml.motion_forecast` to `payload_trajectory_definition.yaml` when present and decide whether the declared path is physically satisfiable; treat a missing or hand-wavy payload trajectory estimate as a strong reject because downstream machinery timing and sizing are then ungrounded; require `plan.md`'s detailed calculation section to fully derive the path-relevant distances, times, speeds, clearances, and tolerances; treat the planner evidence and technical-drawing scripts as the source of the approved benchmark contract; verify exact label, quantity, COTS identity, motion, timing, speed plausibility, and schema grounding; and write the stage-scoped decision/comments YAML pair under reviews/ before submitting the review gate."
Copy file name to clipboardExpand all lines: .agents/skills/benchmark-planner/SKILL.md
+5Lines changed: 5 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -21,6 +21,9 @@ Use the shared preview helpers when evidence or objective overlays are part of t
21
21
-`preview(...)` for live scene inspection and benchmark preview renders
22
22
-`preview_drawing()` for drafting packages and orthographic plan evidence
23
23
-`objectives_geometry()` when a benchmark preview needs reconstructed objective overlays
24
+
-`list_render_bundles()` when the task depends on the exact bundle revision
25
+
-`query_render_bundle()` when you need compact bundle metadata or frame slices
26
+
-`pick_preview_pixel()` / `pick_preview_pixels()` when a preview bundle needs click-to-world evidence
24
27
- Prefer `utils.preview` for new import paths; `utils.visualize` is compatibility-only
25
28
26
29
## Geometry Contract
@@ -79,7 +82,9 @@ Read these before drafting or revising the handoff:
79
82
- Treat `benchmark_plan_technical_drawing_script.py` as display-only: it should not re-author a duplicate shape tree or a second copy of the benchmark geometry, only the orthographic drawing/view scaffolding for the same approved contract.
80
83
- Keep every dimension formula-backed; if the handoff is missing a needed length, thickness, clearance, or placement datum, fix the source rather than guessing.
81
84
- Use `preview(...)` for live scene previews and `preview_drawing()` for drafting packages; they are not interchangeable.
85
+
- Use `payload_path=True` on `preview(...)` when the live payload-path overlay is part of the inspection.
82
86
- When drawings are part of the handoff, inspect the drafted package with `preview_drawing()` before `submit_plan()`.
87
+
- If the review depends on bundle identity or a click-to-world answer, select the exact bundle with `list_render_bundles()` and inspect or query that bundle-local snapshot instead of assuming the newest visible render is the right one.
83
88
- Make the benchmark-owned motion contract explicit if any fixture moves. In this repo, keep each moving fixture to one explicit DOF axis and spell out the controller facts and limits.
84
89
- Keep the moved object inside `build_zone` under static variation plus runtime jitter.
85
90
- Keep goal and forbid zones non-overlapping with the moved object at spawn.
Copy file name to clipboardExpand all lines: .agents/skills/benchmark-reviewer/SKILL.md
+4Lines changed: 4 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -14,6 +14,9 @@ When preview evidence or objective overlays are present, use the shared preview
14
14
-`preview(...)` for live scene previews or benchmark evidence inspection
15
15
-`preview_drawing()` for drafting-package review evidence
16
16
-`objectives_geometry()` when a benchmark preview needs reconstructed objective overlays
17
+
-`list_render_bundles()` when exact bundle identity matters
18
+
-`query_render_bundle()` when you need bundle metadata without the full media payload
19
+
-`pick_preview_pixel()` / `pick_preview_pixels()` when a preview bundle needs click-to-world evidence
17
20
- Prefer `utils.preview` for new code paths; `utils.visualize` is a compatibility alias
18
21
19
22
## Review Checklist
@@ -22,6 +25,7 @@ When preview evidence or objective overlays are present, use the shared preview
22
25
-[ ] Read `benchmark_script.py`, `benchmark_definition.yaml`, `benchmark_assembly_definition.yaml`, `plan.md`, `todo.md`, `validation_results.json`, `simulation_result.json`, `scene.json`, `benchmark_plan_evidence_script.py`, `benchmark_plan_technical_drawing_script.py`, and any `renders/benchmark_renders/**` as read-only context.
23
26
-[ ] Require validation and simulation success for the latest revision, and confirm `goal_reached` in the manifest/result.
24
27
-[ ] Inspect render images with `inspect_media(...)` whenever they exist; if moving benchmark fixtures exist, inspect the latest dynamic simulation evidence before approval.
28
+
-[ ] If bundle identity or a pixel-to-world question matters, resolve the exact bundle with `list_render_bundles()` and query that bundle-local snapshot before approval.
25
29
-[ ] Treat `benchmark_plan_evidence_script.py` and `benchmark_plan_technical_drawing_script.py` as the inspectable source of the approved benchmark contract.
26
30
-[ ] Verify geometric validity, solvability, runtime randomization, exact inventory grounding, and benchmark-side motion against the approved contract and observed evidence. See `references/review_contracts.md`.
27
31
-[ ] Reject implementations that rely on free-form XYZ placement instead of selector-driven placement, explicit mates/joints, or the few absolute anchors already fixed by the approved plan.
short_description: "Review implemented benchmarks after validation and simulation."
4
-
default_prompt: "Use this skill when validating the implemented benchmark after validation and simulation. Read `benchmark_script.py`, `benchmark_definition.yaml`, `benchmark_assembly_definition.yaml`, `validation_results.json`, `simulation_result.json`, `scene.json`, and any benchmark preview media as read-only; follow the checklist in `SKILL.md` and `references/review_contracts.md`; treat the planner evidence and technical-drawing scripts as the source of the approved benchmark contract; verify geometry, solvability, randomization, motion contracts, and inventory exactness; inspect render or video evidence when present; and write only the stage-scoped benchmark-execution review YAML pair under `reviews/` before submitting the review."
4
+
default_prompt: "Use this skill when validating the implemented benchmark after validation and simulation. Read `benchmark_script.py`, `benchmark_definition.yaml`, `benchmark_assembly_definition.yaml`, `validation_results.json`, `simulation_result.json`, `scene.json`, and any benchmark preview media as read-only; follow the checklist in `SKILL.md` and `references/review_contracts.md`; treat the planner evidence and technical-drawing scripts as the source of the approved benchmark contract; verify geometry, solvability, randomization, motion contracts, and inventory exactness; inspect render or video evidence when present, and use list_render_bundles(), query_render_bundle(), or pick_preview_pixel() when exact bundle identity matters; and write only the stage-scoped benchmark-execution review YAML pair under `reviews/` before submitting the review."
Copy file name to clipboardExpand all lines: .agents/skills/engineer-coder/SKILL.md
+20-5Lines changed: 20 additions & 5 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,6 +1,6 @@
1
1
---
2
2
name: engineer-coder
3
-
description: Problemologist engineering implementation role. Use when turning approved engineering handoffs into solution_script.py, solving engineering evals with bounded retries, selecting mechanism patterns, validating and simulating revisions, inspecting render evidence, preserving planner inventory exactness, grounding work in proof-backed plan.md contracts, or refusing an infeasible plan with plan_refusal.md.
3
+
description: Problemologist engineering implementation role. Use when turning approved engineering handoffs into solution_script.py, solving engineering evals with bounded retries, selecting mechanism patterns, validating and simulating revisions, inspecting render evidence, querying render-bundle history or point-pick results, preserving planner inventory exactness, grounding work in proof-backed plan.md contracts, or refusing an infeasible plan with plan_refusal.md.
4
4
---
5
5
6
6
# Engineer Coder
@@ -34,19 +34,31 @@ Use the runtime helpers explicitly in authored engineer scripts:
34
34
35
35
```python
36
36
from utils.submission import validate, simulate
37
-
from utils.preview import preview, preview_drawing, objectives_geometry
37
+
from utils.preview import (
38
+
list_render_bundles,
39
+
objectives_geometry,
40
+
pick_preview_pixel,
41
+
pick_preview_pixels,
42
+
preview,
43
+
preview_drawing,
44
+
query_render_bundle,
45
+
)
38
46
```
39
47
40
48
-`validate(result)` and `simulate(result)` are the required pre-handoff checks.
41
49
-`submit_for_review(result)` is available for supporting scripts that own the final handoff step.
42
-
-`preview(...)` and `preview_drawing()` are the evidence-generation paths; `objectives_geometry()` reconstructs benchmark objective overlays when needed.
50
+
-`preview(...)` is the live scene and objective-overlay path; use `payload_path=True` only when the current workflow needs the live payload-path overlay. `preview_drawing()` is the drafting-package path and keeps the payload overlay off.
51
+
-`objectives_geometry()` reconstructs benchmark objective overlays when needed.
52
+
-`list_render_bundles()` locates the exact current or historical render bundle instead of assuming the latest file on disk is the right snapshot.
53
+
-`query_render_bundle()` returns compact frame/object slices when you need bundle metadata without the full media payload.
54
+
-`pick_preview_pixel()` and `pick_preview_pixels()` resolve screen-space points against the bundle-local snapshot when you need click-to-world or batch point-pick evidence.
43
55
-`from utils.visualize import ...` is a compatibility alias, but `utils.preview` is the preferred namespace for new code.
44
56
45
57
## What This Skill Owns
46
58
47
59
- Implementation strategy for the engineer-coder role.
48
60
- File-level execution discipline for `solution_script.py`, `todo.md`, and `journal.md`.
49
-
- Validation, simulation, media inspection, and review-submission behavior.
61
+
- Validation, simulation, media inspection, render-history lookup, point-pick queries, and review-submission behavior.
50
62
- The decision rules for when to load specialist skills and when to refuse a plan.
51
63
52
64
## What This Skill Does Not Own
@@ -111,7 +123,7 @@ Do not invent fallback behavior to bridge contradictions. If the handoff is inco
111
123
04. Draft the smallest physically credible solution that can survive the declared runtime variation.
112
124
05. Keep `solution_script.py` import-safe and bind the final object as `result`.
113
125
06. Run a cheap syntax/import check, then a real validation probe, then simulation.
114
-
07. Inspect render or video evidence as soon as motion is uncertain. Do not spend extra geometry effort before checking the first failing dynamic result.
126
+
07. Inspect render or video evidence as soon as motion is uncertain. When the question depends on bundle identity or pixel-to-world mapping, use `list_render_bundles()`, `query_render_bundle()`, or `pick_preview_pixel()` against the exact bundle before spending extra geometry effort.
115
127
08. Fix the narrowest failure mode and repeat.
116
128
09. Submit for review only when the latest revision is actually ready.
117
129
10. Anchor any binding numeric claim to the tightened `plan.md` proof structure; do not implement against prose-only assumptions when `ASSUMP-*` and `CALC-*` scaffolding is expected.
@@ -156,6 +168,8 @@ This role should behave like a high-confidence solver, not a wandering explorer.
156
168
- Simulation success is necessary but not sufficient.
157
169
- A passing validator does not excuse skipping the simulation video or frame evidence. Use the first dynamic result to confirm direction, capture, and stability.
158
170
- If `preview(...)` evidence exists for the current revision, inspect the corresponding render bundle before finishing.
171
+
- If render history matters, use `list_render_bundles()` and `query_render_bundle()` to select the exact bundle snapshot instead of assuming the newest visible render is the right one.
172
+
- If a question depends on screen-space picking, use `pick_preview_pixel()` or `pick_preview_pixels()` against that bundle-local snapshot before changing geometry.
159
173
- If render images exist for the current revision, inspect them with `inspect_media(...)` before finishing.
160
174
- If the solution has moving behavior and simulation video exists, inspect the dynamic evidence before approval.
161
175
- Treat `validation_results.json`, `simulation_result.json`, and render manifests as evidence inputs, not as substitutes for reasoning.
@@ -170,6 +184,7 @@ This role should behave like a high-confidence solver, not a wandering explorer.
170
184
5. Robustness failure: widen tolerances to runtime jitter and remove exact-seed dependence.
171
185
6. Directional motion failure: verify slope sign, handedness, and capture path from the actual simulation media before changing more geometry. Also check whether the preview is a front or rear view: preview yaw is clockwise from front, so rear views naturally swap left/right on screen and are not automatically an X-axis mirror.
172
186
7. Reviewer failure: resolve the valid checklist items directly and ignore non-applicable demands.
187
+
8. Render-history or point-pick failure: identify the exact bundle first, then inspect or query that bundle-local snapshot; do not infer coordinates from stale filenames or the latest artifact by default.
173
188
174
189
If the same blocker persists after one targeted fix, record it in `journal.md` and stop widening the search.
0 commit comments