craftassay 0.1.18 → 0.1.19
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/README.md +8 -5
- package/package.json +1 -1
- package/skills/craftassay/SKILL.md +7 -4
- package/skills/craftassay/SKILL_FACTS.md +6 -4
- package/skills/craftassay/references/comparison.md +7 -0
- package/skills/craftassay/references/coverage.md +3 -1
- package/skills/craftassay/references/dimensions.md +9 -0
- package/skills/craftassay/references/report.md +2 -1
- package/skills/craftassay/references/value-clarity.md +31 -0
package/README.md
CHANGED
|
@@ -6,11 +6,14 @@
|
|
|
6
6
|
|
|
7
7
|
An installable review skill for AI agents.
|
|
8
8
|
|
|
9
|
-
|
|
10
|
-
CraftAssay
|
|
11
|
-
|
|
12
|
-
|
|
13
|
-
The report is an assessment, not a certification.
|
|
9
|
+
Know what to improve first. Give your agent a tool, page, or portfolio.
|
|
10
|
+
CraftAssay checks whether its value comes across, weighs the evidence, and
|
|
11
|
+
ranks useful improvements. Findings and scores can be compared across revisions.
|
|
12
|
+
|
|
13
|
+
The report is an assessment, not a certification. It also checks whether a
|
|
14
|
+
newcomer can tell who the tool is for, why its result matters, and how to start.
|
|
15
|
+
The current method is specification v1.2, rubric v1.1. Earlier sample reviews
|
|
16
|
+
retain their original versions.
|
|
14
17
|
|
|
15
18
|
**Get started:** pick the agent, install the skill, then run the
|
|
16
19
|
HarborNote page. Instructions are at
|
package/package.json
CHANGED
|
@@ -30,9 +30,9 @@ sense when it earns its keep. Do not split the product name. Do
|
|
|
30
30
|
not tell the reader they can skip a word in it. Do not lecture
|
|
31
31
|
them on how to capitalize it.
|
|
32
32
|
|
|
33
|
-
The contract is Review and Comparison Specification v1.
|
|
34
|
-
rubric v1.
|
|
35
|
-
(`craftassay-review-spec-v1.
|
|
33
|
+
The contract is Review and Comparison Specification v1.2, scoring
|
|
34
|
+
rubric v1.1. If a full spec file is in the repo
|
|
35
|
+
(`craftassay-review-spec-v1.2.md`), read it when a rule here is
|
|
36
36
|
thin. These reference files win for the report shape you emit.
|
|
37
37
|
|
|
38
38
|
## Review workflow
|
|
@@ -48,7 +48,9 @@ thin. These reference files win for the report shape you emit.
|
|
|
48
48
|
to compare different subjects is allowed.
|
|
49
49
|
3. Choose a **new run folder**. Never write into a folder that
|
|
50
50
|
already has `report.md`. Never overwrite a supplied baseline.
|
|
51
|
-
4. Follow `references/coverage.md`.
|
|
51
|
+
4. Follow `references/coverage.md`. For discovery surfaces, use
|
|
52
|
+
`references/value-clarity.md` before internal specifications.
|
|
53
|
+
Record what you inspected.
|
|
52
54
|
Stop once the evidence is enough for the stated depth.
|
|
53
55
|
5. Write the deliverables using `references/report.md`.
|
|
54
56
|
Create the run folder if needed. If a matching baseline exists,
|
|
@@ -119,4 +121,5 @@ check is clear. That is what `verify-first` is for.
|
|
|
119
121
|
- `references/dimensions.md` — before you assign scores.
|
|
120
122
|
- `references/findings.md` — before you emit a finding.
|
|
121
123
|
- `references/coverage.md` — before you claim what was checked.
|
|
124
|
+
- `references/value-clarity.md` — when reviewing a discovery surface.
|
|
122
125
|
- `references/comparison.md` — when a baseline review exists.
|
|
@@ -2,7 +2,7 @@
|
|
|
2
2
|
skill_facts_version: "0.1.0"
|
|
3
3
|
name: CraftAssay
|
|
4
4
|
developer: Catalyst Forge
|
|
5
|
-
version: 0.1.
|
|
5
|
+
version: 0.1.19
|
|
6
6
|
status: active
|
|
7
7
|
license: MIT
|
|
8
8
|
kind: cursor-skill
|
|
@@ -32,11 +32,13 @@ bundled_artifacts:
|
|
|
32
32
|
kind: other
|
|
33
33
|
- path: references/report.md
|
|
34
34
|
kind: other
|
|
35
|
+
- path: references/value-clarity.md
|
|
36
|
+
kind: other
|
|
35
37
|
egress:
|
|
36
38
|
telemetry: none
|
|
37
39
|
destinations: []
|
|
38
40
|
generated:
|
|
39
|
-
date: 2026-
|
|
41
|
+
date: 2026-10-18
|
|
40
42
|
generator: skillfacts-from-pack
|
|
41
43
|
credits:
|
|
42
44
|
generated_with: https://skillfacts.dev
|
|
@@ -48,7 +50,7 @@ credits:
|
|
|
48
50
|
| | |
|
|
49
51
|
|---|---|
|
|
50
52
|
| **Developer** | Catalyst Forge |
|
|
51
|
-
| **Version** | 0.1.
|
|
53
|
+
| **Version** | 0.1.19 |
|
|
52
54
|
| **Status** | active |
|
|
53
55
|
| **License** | MIT |
|
|
54
56
|
| **Kind** | cursor-skill |
|
|
@@ -96,4 +98,4 @@ Evidence-backed review of usefulness, clarity, quality, and presentation.
|
|
|
96
98
|
---
|
|
97
99
|
*Generated with [SkillFacts](https://skillfacts.dev)*
|
|
98
100
|
|
|
99
|
-
[skillfacts-label]: https://skillfacts.dev/v#sf1.
|
|
101
|
+
[skillfacts-label]: https://skillfacts.dev/v#sf1.eNqdUsFu2zAM_RWDZ9lZboVuQ7EBRbNTdyuCgpFoW4gseSTlIAjy74OMrtmhh2w3QnrvkXx8F1jAbg0knAgsPDL2-lUEz2DA00Ixz8T1AxXjWbT5nnkgMLAQS8gJLHzptt32AQyIohYBC-g0LBUUg6MkVfjH008wcAzJgwVXWDK3cgwxgoG58JxX1LcleEqO2gO6I_mGaQl0anLfFKG-xEQipnEROejZNL8KxrXA5JuZSSgpasipq6qcF0qYHIG9gOTCtYJRdRa72QxBx3LoXJ42f3Zr193a3e5x46oP-O7DXA4xyPiZD1cDIYlycbWtvDGhG9d-I8UIFlJO1YhEesp8vD30IZKcRWkCC0zo2xMHXQU151iVeuLqhQf7ujdwKMlH8m_IGnp0KmBfLzCjjmDh5flpt-smf7M4a533aj4gH3qycXmakYPk9A-UhRgHupvgw0SpBkTupvQh-ZCG-wm56Fz0bjjTnPl--IKxUPuetc9YewM0MInUeytFmkj5fDuxJ9GQ1jzWY-2vBsY80YzD3zm8Ja3ztICBOqYEzavUf6RVuSSHWnOjXOj6GyOrWW4
|
|
@@ -39,3 +39,10 @@ count. Judge whether the underlying user problem was addressed.
|
|
|
39
39
|
|
|
40
40
|
A name change alone must never change a score or imply that a
|
|
41
41
|
finding was resolved.
|
|
42
|
+
|
|
43
|
+
Specification v1.2 / rubric v1.1 adds an explicit value-as-presented check.
|
|
44
|
+
For older baselines, disclose this methodology change and whether equivalent
|
|
45
|
+
public evidence was inspected. Preserve old scores. Attribute changes to
|
|
46
|
+
product evidence, presentation changes, or method changes as supported. If
|
|
47
|
+
these cannot be separated, state that limitation instead of claiming a
|
|
48
|
+
measured improvement. The numerical anchors and seven dimensions remain.
|
|
@@ -4,7 +4,9 @@
|
|
|
4
4
|
|
|
5
5
|
Adapt depth to the subject. Do not skip the start.
|
|
6
6
|
|
|
7
|
-
1. Read the catalog or landing surface as a newcomer.
|
|
7
|
+
1. Read the catalog or landing surface as a newcomer. Apply
|
|
8
|
+
`value-clarity.md` before internal specifications. Keep the seven
|
|
9
|
+
answers, status, source, and reader consequence in `coverage.md`.
|
|
8
10
|
2. Enumerate every in-scope project or included page.
|
|
9
11
|
3. Read each card or primary description completely.
|
|
10
12
|
4. Follow the primary public destination.
|
|
@@ -35,3 +35,12 @@ Overall appeal is not the average of the other six.
|
|
|
35
35
|
|
|
36
36
|
Uniqueness is not a proxy for usefulness. Clarity is comprehension.
|
|
37
37
|
Aesthetics is the effectiveness of the visual presentation.
|
|
38
|
+
|
|
39
|
+
## Rubric v1.1
|
|
40
|
+
|
|
41
|
+
The seven dimensions and numerical anchors are unchanged. The value-as-presented
|
|
42
|
+
check in `value-clarity.md` makes the evidence used for Utility, Clarity,
|
|
43
|
+
Aesthetics, and Accuracy explicit. Distinguish product usefulness from its
|
|
44
|
+
explanation, and reference one finding across affected dimensions without
|
|
45
|
+
counting it as several defects. This evidence procedure is a methodology
|
|
46
|
+
change when comparing against rubric v1.0.
|
|
@@ -6,7 +6,7 @@ Title:
|
|
|
6
6
|
|
|
7
7
|
`CraftAssay Review: <subject>`
|
|
8
8
|
|
|
9
|
-
Then review date, specification version **v1.
|
|
9
|
+
Then review date, specification version **v1.2**, rubric version **v1.1**.
|
|
10
10
|
|
|
11
11
|
For a comparison, use `CraftAssay Comparison: <subject>` and add the
|
|
12
12
|
baseline mapping after the actions.
|
|
@@ -36,6 +36,7 @@ one site. Do not invent portfolio content.
|
|
|
36
36
|
|
|
37
37
|
**Subject identity:** ...
|
|
38
38
|
**User and job:** ...
|
|
39
|
+
**Value as presented:** ...
|
|
39
40
|
**What works and should remain:** ...
|
|
40
41
|
**Review:** ...
|
|
41
42
|
**Material findings:** ...
|
|
@@ -0,0 +1,31 @@
|
|
|
1
|
+
# Value as presented
|
|
2
|
+
|
|
3
|
+
Use this check for discovery surfaces: a public landing page, catalog, README, or equivalent entry point. For a procedure or internal artifact without a discovery surface, record not applicable and explain why. Do not force marketing copy onto an obvious utility or specialist workflow.
|
|
4
|
+
|
|
5
|
+
Before opening internal specifications, reconstruct the answers available to the intended reader. Preserve that first-reader account, then compare it with examples, setup instructions, and implementation evidence where relevant. An author's explanation can identify intent but cannot fill a gap in the public page.
|
|
6
|
+
|
|
7
|
+
| Question | Look for |
|
|
8
|
+
| --- | --- |
|
|
9
|
+
| Why me? | Recognizable person, task, or situation, with realistic expertise and setup requirements |
|
|
10
|
+
| What is it? | Product form and job, such as an agent skill, command-line tool, app, or specification |
|
|
11
|
+
| So what? | The inconvenience, uncertainty, repeated effort, or missed opportunity that makes the job matter |
|
|
12
|
+
| So that? | The decision, action, retained knowledge, or other bounded outcome the result enables |
|
|
13
|
+
| Show me | Representative input, output, and consequence, with illustrations distinguished from observed results |
|
|
14
|
+
| Why this approach? | A supported reason to choose this workflow, including relevant tradeoffs, without invented competitive superiority |
|
|
15
|
+
| What next? | A first action, its prerequisites, and what successful first use produces |
|
|
16
|
+
|
|
17
|
+
In coverage.md, record each answer as **explicit**, **inferable**, **missing**, **contradictory**, or **not applicable**, with its source and reader consequence. These statuses are observations, not scores. Inferable can be sufficient: a visible stopwatch with Start and Reset need not explain why timing a task is useful. Not applicable needs a reason, not a guessed answer.
|
|
18
|
+
|
|
19
|
+
In report.md, write a short **Value as presented** paragraph explaining whether the reader can recognize their situation, understand the bounded result, and choose a next step. State what the product appears to offer separately from how well the page communicates it. Weak explanation does not establish that the product has little value or that market demand is absent.
|
|
20
|
+
|
|
21
|
+
Map the evidence to existing dimensions. Relevance and adoption burden inform Utility. Comprehension informs Clarity. Obscured examples and visual hierarchy inform Aesthetics when the interface was inspected. Unsupported promises inform Accuracy. Record one finding per underlying problem and reference that finding from affected dimensions. Do not multiply priority or severity by the number of dimensions it touches. Use NR when evidence is insufficient.
|
|
22
|
+
|
|
23
|
+
## Calibration
|
|
24
|
+
|
|
25
|
+
- A specialist parser names its input format, output schema, prerequisites, and example. A general visitor's unfamiliarity is not a defect for its stated specialist audience.
|
|
26
|
+
- A timer visibly starts, pauses, and resets. Its purpose can be inferable. Do not demand a slogan, competitor table, or elaborate persona.
|
|
27
|
+
- A polished page promises guaranteed results without evidence. Record a claim-risk finding. Polish does not validate the claim.
|
|
28
|
+
- A useful runnable tool has a vague opening. Identify the missing public explanation under Clarity, and preserve the evidence that the workflow itself is useful.
|
|
29
|
+
- A complete entry page has no material gap. Say so. Do not invent a positioning task to populate the findings register.
|
|
30
|
+
|
|
31
|
+
This check needs no other review skill, proprietary service, or network access when the supplied evidence is local.
|