09.03.05 — Pick skills#

Picking artifacts (screenshots, logs, reports). Always by mtime, never by filename.

Listing (potentially non-exhaustive)#

“mtime”#

Holy Book quote:

Mtime, not filename. UUID suffixes are random; pick-* sorts by mtime.

All Ocarina artifacts are named with a UUID:

• Screenshots: SUCCESS_a3f2b1c4.png, FAIL_b9e8d2a3.png
• Logs: <test name>.log but in a <cycle>/<campaign>/<suite>/<test>.log tree, often duplicated between runs with a root folder suffixed by a UUID for multiple runs on the same machine
• Reports: <uuid hex 8>.json, <uuid hex 4>/<...>.docx

Schema#

input  : "run review-report on the last run"
        → pick-reports (mtime DESC, limit 1)
        → review-report on the picked report

input  : "find screenshots of test X from the last run"
        → pick-reports → last run
        → pick-screenshots scoped

Why dedicated skills rather than ls -t#

“pick then act”#

Failing cycle  →  pick-reports  →  review-report(<picked report>)
                                            ↓
                                    classification → IDENTIFIED_GAPS.md

pick-* is rarely called alone. It’s always a prelude to a skill that consumes the picked artifact.

Random-naming illustration#

When an Ocarina-driven test cycle finishes, artifacts are built like:

.reports/
├── docx/
│   └── <uuid 4 chars>/                                           # per run
│       └── <campaign>/
│           └── <suite>/
│               └── <test>.docx
└── json/
    └── <uuid 8 chars>.json                                       # per run