kklimuk/docx-cli / .claude/skills/weak-agent-test
weak-agent-test
VerifiedRun the weak-agent adversarial test harness against docx-cli. Spawns weak exercise agents (Haiku by default, Sonnet to probe, or a local agent harness's pre-produced runs) to perform real document tasks over six scenarios — five editing (MNDA form-fill + font fidelity, invoice table-edit/restructure + logo replace, résumé styling, contract redlining + commenting, contract finalize via accept/reject + comment reply/resolve) and one authoring (T. S. Eliot poetry journal: multi-column, verse, footnotes, links, figure) — renders every result with Word, has opus judge them against ground-truth rubrics, measures each exercise's tool economy, token cost, wall-clock, and correctness (from transcripts for Claude, the exercise.json ledger for the local harness), and synthesizes a prioritized ergonomics report. Use when the user says 'adversarial review', 'test docx-cli with weak agents', 'run the haiku harness', 'weak agent test', or wants to re-run yesterday's adversarial process.
scenarios/mnda/criteria.md
View on GitHub# Grading rubric — mnda (JUDGE ONLY)
The agent never sees this file. It is the ground-truth definition of "correct" for
the request in `task.md`. Judge the finished document against these checks.
## Pass conditions
- **Cover-page placeholders filled** with the values from `task.md`: both party
names, the signatories + titles, notice emails, the purpose, Effective Date
(June 8, 2026), MNDA term (2 years), confidentiality term (3 years), Delaware
governing law, New Castle County jurisdiction, and the signature-block dates. No
`[ ... ]` bracket placeholders remain anywhere.
- **Signature table populated in the intended blank cells** for BOTH parties — in
table `t2`, Party 1 values occupy `c1` and Party 2 values occupy `c3`: Print Name
(`r2`), Title (`r3`), Company (`r4`), Notice Address (`r5`), and Date (`r6`).
The narrow spacer column `c2`, label column `c0`, and blank signature row `r1`
remain untouched. Each destination stays a single paragraph (`p0`); extra `p1`
lines or values placed in neighboring cells fail.
- **Zero yellow highlight** anywhere in the document. The placeholders were
highlighted; once filled, no highlight may remain (check the runs, not just the
visual — a filled value that kept its `<w:highlight>` fails).
- **Font fidelity preserved** — the Georgia title, the bold field labels, the small
gray sub-labels, and the Arial body all keep their original font, size, and color.
The before/after render is identical EVERYWHERE except the filled-in values and the
removed highlight. Any font/size/color drift on untouched text is a failure.
## How to verify
- `docx read FILE` (and `--ast` for run-level detail) — confirm the values landed and
no `[...]` or highlight survives. The direct form-fill path is `docx edit FILE --at
CELL --text ...` against bare cells `t2:r2c1` through `t2:r6c1` and their `c3`
counterparts; explicit `:p0` is an equivalent fallback, but the result must still
have exactly one paragraph per destination cell.
- Compare the BASELINE render against the OUTPUT render: only the intended cells
changed; all other formatting is intact.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |