/check-deliverable — standalone qa gate
Run the existing qa-reviewer agent against ONE finished text deliverable on demand, without a full Ann/Vi orchestration. Reuse qa-reviewer wholesale: add no new checking logic, rewrite nothing, write nothing to disk.
Lane — when this skill, not another
- /check-deliverable (this): checks an existing finished file against the qa gate. One file, one qa-reviewer spawn.
- /ann: produces a deliverable through the full pipeline. Use when the goal is the artefact.
- /ane-voice: audits and rewrites prose voice. Use to fix tone, not to run the full gate.
- /code-review: reviews code diffs, not MEL prose deliverables.
Arguments
/check-deliverable <file> [--tier 1|2] [--subgroup colleague|ma-staff|partner-ngo|management|junior-mel|peer-review]
<file> (required): path to the deliverable.--tier (default 1).--subgroup (default colleague).
Derive voice positioning from the subgroup:
management to directivejunior-mel to collaborative-pedagogical- all other subgroups to
collaborative
Steps
Validate the file. Resolve the path. If the extension is not .md, .txt, or .markdown, STOP and return exactly this, then spawn nothing:
qa-reviewer can only inspect text. For Word, Excel, or PowerPoint deliverables run the full /ann pipeline so the brand and format checks apply.
Spawn qa-reviewer via the Agent tool with subagent_type: "qa-reviewer". Pass this prompt, substituting the bracketed values:
## Standing instructions
Audience tier: [Tier 1|Tier 2]; subgroup: [subgroup]; voice positioning: [positioning]
## Task
Standalone single-file review. Review the deliverable at this path:
[absolute file path]
There is NO Ann plan, NO Definition-of-done criteria, and NO P1 wiki block.
Cold-load the wiki per your session-start fallback (qa-block-schema.md,
domain-standards.md, calibration.md) and omit the Definition-of-done subsection.
Read and Grep the file directly to run every mandatory check (tier register,
em-dash, citation, lens, power-shift, edge-case probe). Populate the qa_block,
add reconciliation notes, and close with your VERDICT line.
Reformat the result BLUF-first:
- Line 1: the verdict (
APPROVED / PASS_WITH_GAPS / REJECTED — reason).
- Then the load-bearing findings in plain prose: hard FAILs first, then gaps, then soft flags.
- Then the full verbatim qa_block under a
### Full qa_block heading.
Reformat only. Never override, soften, or reinterpret a qa-reviewer finding.
Boundaries
- One file, one qa-reviewer spawn. No reader-position-reviewer.
- No prose rewriting (that is
/ane-voice).
- No writes to the deliverable or anywhere on disk.
1---2name: check-deliverable3description: Run the qa-reviewer maker-checker gate against one finished text deliverable without a full Ann orchestration: spawns qa-reviewer alone on a single markdown or text file, sets the audience tier (default Tier 1 / colleague; override with --tier and --subgroup), and returns a BLUF verdict plus the full qa_block. Read-only: rewrites nothing, writes nothing to disk. Use on '/check-deliverable <file>' or any ask to QA a finished brief on its own. Refuses Office binaries (.docx/.xlsx/.pptx). Not for producing a deliverable (/ann), voice rewriting (/ane-voice), or code review (/code-review).4---5
6# /check-deliverable — standalone qa gate
7
8Run the existing **qa-reviewer** agent against ONE finished text deliverable on demand, without a full Ann/Vi orchestration. Reuse qa-reviewer wholesale: add no new checking logic, rewrite nothing, write nothing to disk.
9
10## Lane — when this skill, not another
11- **/check-deliverable** (this): *checks* an existing finished file against the qa gate. One file, one qa-reviewer spawn.
12- **/ann**: *produces* a deliverable through the full pipeline. Use when the goal is the artefact.
13- **/ane-voice**: audits and rewrites prose voice. Use to fix tone, not to run the full gate.
14- **/code-review**: reviews code diffs, not MEL prose deliverables.
15
16## Arguments
17
18`/check-deliverable <file> [--tier 1|2] [--subgroup colleague|ma-staff|partner-ngo|management|junior-mel|peer-review]`
19
20- `<file>` (required): path to the deliverable.
21- `--tier` (default `1`).
22- `--subgroup` (default `colleague`).
23
24Derive voice positioning from the subgroup:
25- `management` to `directive`
26- `junior-mel` to `collaborative-pedagogical`
27- all other subgroups to `collaborative`
28
29## Steps
30
311. **Validate the file.** Resolve the path. If the extension is not `.md`, `.txt`, or `.markdown`, STOP and return exactly this, then spawn nothing:
32
33 > qa-reviewer can only inspect text. For Word, Excel, or PowerPoint deliverables run the full `/ann` pipeline so the brand and format checks apply.
34
352. **Spawn qa-reviewer** via the Agent tool with `subagent_type: "qa-reviewer"`. Pass this prompt, substituting the bracketed values:
36
37 ```
38 ## Standing instructions
39 Audience tier: [Tier 1|Tier 2]; subgroup: [subgroup]; voice positioning: [positioning]
40
41 ## Task
42 Standalone single-file review. Review the deliverable at this path:
43 [absolute file path]
44
45 There is NO Ann plan, NO Definition-of-done criteria, and NO P1 wiki block.
46 Cold-load the wiki per your session-start fallback (qa-block-schema.md,
47 domain-standards.md, calibration.md) and omit the Definition-of-done subsection.
48 Read and Grep the file directly to run every mandatory check (tier register,
49 em-dash, citation, lens, power-shift, edge-case probe). Populate the qa_block,
50 add reconciliation notes, and close with your VERDICT line.
51 ```
52
533. **Reformat the result BLUF-first:**
54 - Line 1: the verdict (`APPROVED` / `PASS_WITH_GAPS` / `REJECTED — reason`).
55 - Then the load-bearing findings in plain prose: hard FAILs first, then gaps, then soft flags.
56 - Then the full verbatim qa_block under a `### Full qa_block` heading.
57
58 Reformat only. Never override, soften, or reinterpret a qa-reviewer finding.
59
60## Boundaries
61- One file, one qa-reviewer spawn. No reader-position-reviewer.
62- No prose rewriting (that is `/ane-voice`).
63- No writes to the deliverable or anywhere on disk.
gasserane