feat: three-tier agent skills — users, contributors, maintainers (PTC-80) - #196
Open
db-tycoon-stephen wants to merge 2 commits into
Open
feat: three-tier agent skills — users, contributors, maintainers (PTC-80)#196db-tycoon-stephen wants to merge 2 commits into
db-tycoon-stephen wants to merge 2 commits into
Conversation
…-80) Seed skills/ so coding agents can drive the tycoon CLI correctly: four user-tier skills (adding sources, diagnosing projects, reading state, building models/dashboards), two contributor-tier (branch model, test layers), one maintainer-tier (release cycle). Distributed as three plugins via .claude-plugin/marketplace.json rather than .claude/skills/ because Claude Code does not discover nested skill folders, and user-tier consumers never clone the repo. Every command claim was verified against --help output, and each user-tier skill was tested baseline-vs-with-skill on scripted scenarios (the baselines wrongly trusted tycoon doctor's exit code and misplaced the raw database — both now covered). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Skills align 1:1 with command families rather than individual commands: --help already covers single-command mechanics, and the valuable knowledge sits between the commands of a family. The README index makes coverage auditable — every command maps to an owning skill or explicitly to none, and unmapped commands are treated as undocumented. operating-pipelines covers the unattended-operations family (run-all, schedule, notify, sync, clean). Baseline testing showed this family's help text already answers most questions; the skill's value is consolidation (one read vs eight help calls) plus the cross-family traps (run-all profile resolution, severity filtering). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Jira: PTC-80
First stab at the agent-skills subdirectory discussed in the 7/28 sync — skills so coding agents (Claude Code, Codex, Cursor, …) can drive the tycoon CLI correctly, for three audiences.
What's here
Design decisions (details in skills/README.md)
.claude/skills/. Claude Code doesn't discover nested folders under.claude/skills/(upstream issues #28266/#18192), and user-tier consumers never clone this repo — so the tiers ship as three installable plugins from a top-levelskills/tree, the same pattern dbt-labs/duckdb/astral use./plugin marketplace add Database-Tycoon/tycoon-cli.--helpalready covers single-command mechanics; the skills carry what sits between commands (the add→run→analyze→transform arc,doctorvsprofiles doctorexit codes,run-all's hardcoded profile resolution). The README index maps every command to its owning skill — or explicitly to none — so coverage stays auditable.How it was verified
Every command/flag claim checked against live
--helpon this branch. Each user-tier skill was scenario-tested: a fresh agent answered realistic asks without the skill (baseline) and with it. Baselines produced real failures the skills now cover — trustingtycoon doctor's exit code as a CI gate (it always exits 0), placing raw data in per-source DB files, inventing acatalog installsubcommand from a stale hint. Full suite green (706 passed); the new files are outside every test-harness scan path.Open questions for review
tycoon initadditionally scaffold theusers/tier into new projects? Two traps noted in skills/README.md if so:scaffold_from_template()substitutes{{ }}in.mdfiles, andtycoon-test:-marked bash blocks get executed by the recipe doctest harness.🤖 Generated with Claude Code
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is disabled.