How do I install the Selection Report skill?

Run npx skillmds@latest add jperrello/selection-report in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Selection Report safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Selection Report?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Selection Report free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Selection Report?

jperrello (@jperrello) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Selection Report by jperrello — Agent Skill

Name: Selection Report
Author: jperrello

Selection Report

Write output/<slug>/_selection.json for a source — the shipped shorts (scores + rationale from segments.raw.json, linked to their delivered filename) ALONGSIDE the considered-not-shipped RLM candidate menu (candidates.hint.json, each marked picked/unused) and the topic list (topics.json). Deterministic, no Claude, idempotent, non-fatal. Runs once at end of run AND standalone over the whole output/ backlog. Answers "show me the other arguments alongside the shorts".

jperrello 25117b5 3 files · 8.5 KB Updated Sep 17, 2026

File contents

selection-report

A read-only audit artifact: for each source, what got shipped, what was considered and passed over, and why. Sources every field from work/<id>/ — no Claude.

Usage

selection-report.sh <work_dir> [output_root]   # one source -> output/<slug>/_selection.json
selection-report.sh --backlog [output_root]    # every work/<id> with a matching output/<slug>/

output_root defaults to $OUTPUT_DIR (.env) or <repo>/output.

Inputs (all from `work/<id>/`, all optional → graceful empty sections)

ingest.json — title/url/duration + the slug that names output/<slug>/.
segments.raw.json — the produced shorts with scores + rationale (the SHIPPED set).
clip_NN.done.completion — written by start.sh save-local; maps span N → the delivered filename (the title differs from title_suggestion because generate-title re-titles). Absent in the legacy shorts.sh path → name: null.
candidates.hint.json — the rlm discovery menu (CONSIDERED set; absent on non-rlm short sources → empty).
topics.json — the topic chapters.

Output `output/<slug>/_selection.json`

{
  "source_id": "...", "title": "...", "url": "...", "slug": "...", "duration_sec": 0,
  "shipped_count": 0, "considered_count": 0, "considered_picked": 0, "topics_count": 0,
  "shipped":   [ { "rank", "name", "delivered", "t0","t1","cuts", "hook_type",
                   "opening_line","title_suggestion","rationale","topic",
                   "overall_score","hook_score","context_score","structure_score",
                   "hook_payoff_coherence","payoff_offset_sec","replay_quotient",
                   "thread?","thread_kind?" } ],
  "considered":[ { "t0","t1","quote","why","confidence",
                   "picked","picked_by",            // picked = overlaps a shipped cut
                   "thread?","kind?","cuts?","bridge?" } ],
  "topics":    [ { "t0","t1","title","summary" } ]
}

A candidate is picked: true when its [t0,t1] overlaps any shipped span's cuts; picked_by names the short it overlaps most.

Where it runs

start.sh end-of-run (after sources-ledger, before schedule-drip) for the just-run source; shorts.sh end-of-run; standalone --backlog to backfill existing output dirs. Idempotent — rewritten each run.

---name: selection-reportdescription: Write output/<slug>/_selection.json for a source — the shipped shorts (scores + rationale from segments.raw.json, linked to their delivered filename) ALONGSIDE the considered-not-shipped RLM candidate menu (candidates.hint.json, each marked picked/unused) and the topic list (topics.json). Deterministic, no Claude, idempotent, non-fatal. Runs once at end of run AND standalone over the whole output/ backlog. Answers "show me the other arguments alongside the shorts".---# selection-reportA read-only audit artifact: for each source, what got shipped, what was consideredand passed over, and why. Sources every field from `work/<id>/` — no Claude.## Usage```bashselection-report.sh <work_dir> [output_root] # one source -> output/<slug>/_selection.jsonselection-report.sh --backlog [output_root] # every work/<id> with a matching output/<slug>/````output_root` defaults to `$OUTPUT_DIR` (`.env`) or `<repo>/output`.## Inputs (all from `work/<id>/`, all optional → graceful empty sections)- `ingest.json` — title/url/duration + the slug that names `output/<slug>/`.- `segments.raw.json` — the produced shorts with scores + rationale (the SHIPPED set).- `clip_NN.done.completion` — written by `start.sh` save-local; maps span N → the delivered filename (the title differs from `title_suggestion` because `generate-title` re-titles). Absent in the legacy `shorts.sh` path → `name: null`.- `candidates.hint.json` — the rlm discovery menu (CONSIDERED set; absent on non-rlm short sources → empty).- `topics.json` — the topic chapters.## Output `output/<slug>/_selection.json````json{ "source_id": "...", "title": "...", "url": "...", "slug": "...", "duration_sec": 0, "shipped_count": 0, "considered_count": 0, "considered_picked": 0, "topics_count": 0, "shipped": [ { "rank", "name", "delivered", "t0","t1","cuts", "hook_type", "opening_line","title_suggestion","rationale","topic", "overall_score","hook_score","context_score","structure_score", "hook_payoff_coherence","payoff_offset_sec","replay_quotient", "thread?","thread_kind?" } ], "considered":[ { "t0","t1","quote","why","confidence", "picked","picked_by", // picked = overlaps a shipped cut "thread?","kind?","cuts?","bridge?" } ], "topics": [ { "t0","t1","title","summary" } ]}```A candidate is `picked: true` when its `[t0,t1]` overlaps any shipped span's cuts;`picked_by` names the short it overlaps most.## Where it runs`start.sh` end-of-run (after `sources-ledger`, before `schedule-drip`) for the just-runsource; `shorts.sh` end-of-run; standalone `--backlog` to backfill existing output dirs.Idempotent — rewritten each run.

Frequently asked questions