# AI use cases for Skillranker (Dicklesworthstone/skillranker) Repo: [Dicklesworthstone/skillranker](https://github.com/Dicklesworthstone/skillranker) | Use cases: 5 | Page: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/ SkillRanker is a Rust CLI (`sr`) that uses a Jev-compatible provider to rank agent skills for the next step, based on live session context. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Suggest the right agent skill for each Claude Code step: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/rank-skills-for-next-claude-code-step/ - Inspect a labeled skill ranking offline first: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/try-skillranker-offline-before-connecting/ - Find out why an expected skill was excluded: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/explain-why-a-skill-was-not-suggested/ - Review skill adoption and find coverage gaps: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/review-skill-usage-and-coverage-gaps/ - Replay a labeled evaluation set without network calls: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/replay-labeled-evaluations-offline/ ## How to suggest the right agent skill for each Claude Code step with Skillranker Repo: [Dicklesworthstone/skillranker](https://github.com/Dicklesworthstone/skillranker) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Connect a Claude Code session to SkillRanker so each next step gets a ranked, advisory skill suggestion. **Who it is for:** A developer using Claude Code on a large skill library who keeps losing time picking which procedure fits the current request. **The problem:** A big skill library makes choosing a procedure a task of its own, and a plausible but unsuitable skill can waste context and steer the agent wrong. **The outcome:** Claude Code hooks request a recommendation for the current session, and failed hook calls fall back quietly without blocking the agent. ### Steps 1. Install `sr` from your checkout with `cargo install --locked --path . --bin sr`, or use the installer described in the README. 2. Create a `.env` file from the example, run `chmod 600 .env`, add your own TypeSafe API key, and export the values with `set -a`, `. ./.env`, and `set +a`. 3. Preview the Claude hook settings change with `sr install-hook claude`. 4. Apply the hook change with `sr install-hook claude --apply`. 5. Check a live ranking for the selected session with `sr rank --allow-network`. ### Prompt to paste into your agent ```text Using the Dicklesworthstone/skillranker repo, preview the Claude Code hook change with `sr install-hook claude` for the project at , then apply it only after I confirm. Expected output: the proposed settings diff, then a confirmation that the hook was applied. Verify that `sr doctor --json` reports the hook as configured and that my TYPESAFE_API_KEY is set without printing its value. ``` ### Caveat Fresh rankings need your own provider credential and network consent, and the README says the recommendation is advisory, so the agent still decides what to consult. Risk, credentials: Your provider API key lives in a local .env file, so restrict it with chmod 600 and keep it out of shell history and version control. Risk, remote script: If you use the curl-piped installer, review the script first, since it downloads and runs code on your machine. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/rank-skills-for-next-claude-code-step/ ## How to inspect a labeled skill ranking offline first with Skillranker Repo: [Dicklesworthstone/skillranker](https://github.com/Dicklesworthstone/skillranker) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Run the built-in fixture demo and readiness checks before sharing any real session. **Who it is for:** A cautious developer or team lead who wants to see how SkillRanker behaves before giving it any private conversation data. **The problem:** Connecting a live session to a ranking service means sharing context, and you want to understand the output format and setup gaps first. **The outcome:** You see a labeled fixture result, confirm which adapters and integrations your build supports, and get a dry-run preview of the redacted request with no network calls. ### Steps 1. Install `sr` with `cargo install --locked --path . --bin sr`. 2. Run `sr demo --case useful` to see a labeled fixture result without a key, network, or private session. 3. Run `sr doctor --json` and `sr capabilities --json` to check configuration and available adapters. 4. Run `sr roster --json` to inspect visible, shadowed, and excluded skill records. 5. Preview the redacted wide-pass request with `sr rank --context scratch/context.json --dry-run`. ### Prompt to paste into your agent ```text Using the Dicklesworthstone/skillranker repo, run `sr demo --case useful` and `sr rank --context --dry-run`. Expected output: a labeled fixture result and a redacted request preview with no network call. Verify that the preview contains no API key or secret values before sharing it further. ``` ### Caveat The demo uses an offline fixture, so it shows the result format and not how the ranking performs on your own skill library. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/try-skillranker-offline-before-connecting/ ## How to find out why an expected skill was excluded with Skillranker Repo: [Dicklesworthstone/skillranker](https://github.com/Dicklesworthstone/skillranker) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Trace where a specific skill dropped out of a ranking and what threshold or setting caused it. **Who it is for:** A skill author or agent operator who expected a particular skill to be suggested and got nothing back. **The problem:** When a skill that should have appeared is missing, it is hard to tell whether visibility, retrieval, the abstention rules, or a threshold removed it. **The outcome:** A trace shows the stage where the candidate was excluded, along with thresholds and recovery hints. ### Steps 1. Install `sr` with `cargo install --locked --path . --bin sr`. 2. List the visible, shadowed, and excluded skill records with `sr roster --json`. 3. Run the trace with `sr rank --context scratch/context.json --allow-network --why-not SKILL_ID --explain`, replacing SKILL_ID with the skill you expected. 4. Read the exclusion stage, threshold, and recovery hint in the output. ### Prompt to paste into your agent ```text Using the Dicklesworthstone/skillranker repo, run `sr roster --json` and then `sr rank --context --allow-network --why-not --explain`. Expected output: the stage where was excluded, with thresholds and recovery hints. Verify that the trace reports no extra inference calls beyond the original ranking, as the README describes. ``` ### Caveat The trace explains exclusions in the local pipeline, but the README notes that model-generated reasons are not invented, so you get computed evidence rather than a narrative explanation. Risk, credentials: The live trace needs your own provider credential and sends the selected context over the network after you pass --allow-network. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/explain-why-a-skill-was-not-suggested/ ## How to review skill adoption and find coverage gaps with Skillranker Repo: [Dicklesworthstone/skillranker](https://github.com/Dicklesworthstone/skillranker) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Check which skills are loaded, where descriptions are weak, and where the library lacks coverage. **Who it is for:** A team maintaining a shared skill library who wants to prune, rewrite, or add skills based on local observations. **The problem:** Without data, it is unclear which skills agents actually load, which descriptions are too vague to match, and which tasks have no skill at all. **The outcome:** A local report shows per-skill adoption, description quality issues, and suspected coverage gaps, with the limits of adoption data stated. ### Steps 1. Install `sr` with `cargo install --locked --path . --bin sr`. 2. Review observations with `sr stats --since 7d --by-skill`. 3. Inspect description quality with `sr doctor --descriptions`. 4. List suspected coverage gaps with `sr gaps`. ### Prompt to paste into your agent ```text Using the Dicklesworthstone/skillranker repo, run `sr stats --since 7d --by-skill`, `sr doctor --descriptions`, and `sr gaps`. Expected output: per-skill adoption counts, description quality notes, and suspected coverage gaps. Verify that the report labels adoption as observation and does not present it as proof of usefulness. ``` ### Caveat The README says adoption is not proof of usefulness, so treat these numbers as signals for review rather than measured skill quality. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/review-skill-usage-and-coverage-gaps/ ## How to replay a labeled evaluation set without network calls with Skillranker Repo: [Dicklesworthstone/skillranker](https://github.com/Dicklesworthstone/skillranker) | Difficulty: Advanced | Time: 60 minutes | Basis: From the README Re-run a labeled ranking dataset locally to compare policies without spending provider calls. **Who it is for:** A maintainer tuning SkillRanker policies who needs repeatable results on a fixed labeled dataset. **The problem:** Changing ranking settings and re-querying a live provider is slow, costly, and hard to reproduce. **The outcome:** Offline replay shows how the labeled artifact scores and explains each decision, with no network requests. ### Steps 1. Install `sr` with `cargo install --locked --path . --bin sr`. 2. Prepare a labeled evaluation artifact at `scratch/evaluation.json`, using the case capture described in the README. 3. Replay it with `sr eval --dataset scratch/evaluation.json --explain`. 4. Compare the explained decisions against your expected labels. ### Prompt to paste into your agent ```text Using the Dicklesworthstone/skillranker repo, run `sr eval --dataset --explain`. Expected output: replayed decisions with explanations and no network requests. Verify that the replay reports no provider calls and that the labels in the dataset match the outcomes you expect. ``` ### Caveat The README asks for a labeled evaluation artifact, so you must build or capture the dataset yourself before replay gives useful results. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dicklesworthstone-skillranker/replay-labeled-evaluations-offline/