Pass: 47/50. Failures: edge case prompt parsing. Acceptable for v1.
Pick the one capability the Skill adds
30-60 min
Scaffold the skill folder structure
20 min
Pass: 96/100. New regression test case from real user feedback.
Pass: 145/150. New regression: 3 cases of malformed JSON output. Fix in v1.2.1.
Write SKILL.md: the trigger description is everything
1-2 hr
Add resources: templates, schemas, reference docs
1-3 hr
JSON output bug fixed. Pass: 150/150.
Includes 50 new edge cases from user feedback.
Add scripts: sandbox code the skill executes
2-6 hr (most of it making the script idempotent + handling errors)
Eval the trigger description against real queries
2-4 hr per iteration, 2-5 iterations typical
Package the skill: README, version, install instructions
1-2 hr
Distribute: registry entry, share link, or org install
1-2 hr
Iterate: monitor usage, log regressions, ship v1.1
Ongoing, 1-2 hr/month