Discover Reusable Workflows from Prompt History

This skill reads opencode prompt history, finds repeated multi-step workflows, and recommends candidates worth turning into reusable skills. Invoked via /skill skill-suggester (optionally scoped with --since <date>), it locates prompt history JSONL files at ~/.local/state/opencode/prompt-history*.jsonl, parses each entry, and scores skill potential by looking for repetition (similar phrasing or topic used 3+ times), multi-step sequences (requests needing 5+ tool calls), unsupported requests without a dedicated skill yet, and workaround patterns given manually every time instead of a one-shot command. For each candidate it notes how many times the pattern appeared, how many tool calls it typically consumed, and an estimated time savings if it became a skill. Output is a ranked candidate list, each with the underlying pattern, frequency, average complexity, estimated weekly time savings, direct quoted evidence from past prompts, a priority rating (high = clear ROI used weekly, medium = nice to have, low = rare but worth noting), and a recommendation to create as a skill, add as a command template, or skip as not worth it. It only flags patterns appearing more than twice - one-off requests are never proposed as skills - and if nothing qualifies it says so explicitly rather than inventing a candidate.

Use it when mining opencode prompt history for repeated workflows, recurring unmet needs, or candidates for new reusable skills. Its recommendations are suggestions only, still requiring human review before actually creating or publishing a new skill, and since prompt history can contain sensitive local context, it summarizes patterns rather than exposing unnecessary private excerpts.

Inputs: local opencode prompt history files, optionally scoped by a --since <date> window.
Outputs: a ranked list of skill candidates, each with pattern description, frequency, tool-call complexity, estimated time savings, quoted evidence, priority rating, and a create/template/skip recommendation.

Opencode users who want to discover which of their repeated manual workflows are worth converting into a dedicated skill or command template, rather than re-explaining the same multi-step process every time.