Menu
SkillsMP has collected 34 skills from deephaven/deephaven-mcp. Open a skill to review its source and details.
Showing 34 of 34 collected skills.
Add a new command to the dhcli CLI — invoke when adding or editing a verb under cli/_commands/. Wraps the click + Pattern B + structured-error + agents-manifest conventions; prevents the most common bugs (calling asyncio.run inline, printing errors to stderr,…
The content contract for dhcli CLI help — required sections, HelpSpec authoring, plain-text rendering rules, single-source output schema, and docs/CLI.md consistency — invoke when authoring or reviewing any command help, option help, or the agents manifest
Conventions for dhcli commands that wrap an MCP tool — wrapper categories, type scoping, path-flag locality (CLI-read vs server-side), and the wraps_tool drift contract — invoke when adding/editing a cli/_commands verb that fronts a tool; not for authoring…
The role — audience and scope — of every top-level markdown document in this project. Invoke before editing any top-level doc, and when deciding where new documentation content belongs, to keep edits in-scope and prevent content drift between documents
Reference guide for this project — architecture, server commands and ports, config layout, code quality check commands, test clients — invoke when working with server configuration, running checks, or navigating the codebase
Python coding conventions and style guide for this project — invoke before writing or reviewing any Python under src/ or tests/, including docstring-only edits. Numbered rules covering private-access boundaries, test-file naming and placement, type-hint…
Comprehensively improve unit tests — invoke when a source file's coverage is below the 100%-per-file target, or its test file is missing or disorganized. Ensures test files exist, adds missing tests, removes unneeded ones, and restructures to reach full…
Run the dependency freshness check — invoke before a release, or after changing dependencies in pyproject.toml. Performs a fresh resolve then runs mypy and pytest against it, and diagnoses any failures
Verify dhcli CLI help is factually accurate against the code — invoke for surgical correctness-only fixes when a command's help structure is already complete; use cli-help-improve for a full review. Checks that documented flags, arguments, error codes, exit…
Comprehensively improve the help surface of one or more dhcli CLI commands — fill missing sections, single-source the output schema, fix inaccuracies, and reconcile the description surfaces (the HelpSpec behind --help and the agents manifest, and docs/CLI.md)…
Add a new configuration field, setting, or tunable — invoke when adding a knob to server.json, cli.json, community/, or enterprise/. Covers the Pydantic schema, the PEP 257 trailing docstring, config-samples, and tests; prevents ad-hoc os.environ reads and…
Verify a markdown documentation file is factually accurate — invoke for surgical correctness-only fixes when the document's structure is already sound; use docs-improve for a full review. Checks commands, file paths, config keys, API names, code examples, and…
Comprehensively improve a markdown documentation file — invoke when a doc is thin, disorganized, or stale and a full pass is wanted; use docs-accuracy for correctness-only fixes. Fills missing content, fixes inaccuracies and broken links, reorganizes sections…
Run the integration tests (Docker / pip / subprocess-based) — invoke when verifying a change that touches session launching, packaging, or subprocess handling, since these tests are excluded from the default run. Encodes the required -s flag and the other…
Stress test a deephaven-docs MCP server — invoke when checking a dev or prod docs server for timeouts, rate limiting, or response variation under sustained load. Calls docs_chat N times sequentially (default N=100) and reports pass/fail counts
Add a new MCP tool to the systems server — invoke when adding a tool under mcp_systems_server/_tools/. Wraps ref-mcp-module-organization, pydocs-improve, and ref-logging-standards; prevents the most common bug (forgetting to register the tool)
Verify Python docstrings are factually accurate — invoke for surgical correctness-only fixes during a code review, where restructuring would expand scope; use pydocs-improve for a full pass. Checks descriptions, Args, Returns, and Raises against the actual…
Comprehensively improve Python docstrings — invoke when a file's docstrings are thin, stale, or missing required sections; use pydocs-accuracy for correctness-only fixes. Corrects inaccuracies, adds missing sections, and enforces the required MCP tool…
Standards for editing the project root `AGENTS.md` — what belongs there vs in a skill, format conventions, composition with skills, sync rules — invoke when adding, removing, or restructuring `AGENTS.md` content
Ground dhcli design decisions in the conventions of comparable CLIs — invoke when choosing a noun, verb, flag name, default, or interaction model, or when reviewing one. Requires citing the tool and the specific behavior being followed, and stating the reason…
How Deephaven MCP servers consume runtime configuration — JSON5 + Pydantic v2 + templating, schema defaults, no ad-hoc env reads — invoke before adding a new tunable, refactoring a config model, or wiring an environment variable into the code
Python logging conventions — module-level _LOGGER instantiation, message format ([module:function] Action: details), log levels, sensitive-data redaction, and coverage rules — invoke when writing or reviewing logging statements in Python code
Documentation format standards for markdown files — JSON/JSON5 code block requirements, placeholder formatting, copy-paste readiness — invoke when editing or creating markdown documentation
Module organization and design patterns for MCP tool development in this project — invoke when creating or modifying MCP tool modules
Project conventions for serializing values into user-facing output — MCP tool return dicts and CLI output fields — invoke when authoring or reviewing any string field or payload shape that ships to a user or agent (a return payload or a CLI OutputField)
Structural standards for agent skill files — invoke when adding, editing, or reviewing a skill's structure. Covers the published-guidance baseline, composition, frontmatter, the ref- naming rule, degrees of freedom, body shape, multi-file layout, and…
Content-quality standards for agent-facing prose — invoke when authoring or reviewing anything an AI agent reads at decision or execution time: skill descriptions and bodies, AGENTS.md rules, MCP tool docstrings, error-code help text. Covers triggerability,…
Perform a deep review of a changeset — invoke when reviewing a set of code changes spanning more than one file, before committing; use review-python-file for a single file in depth. Routes each file to its type-specific review skill, then adds the…
Perform a comprehensive review of a single Python file — invoke when reviewing one file in depth; use review-changes for a multi-file changeset. Covers design, correctness, DRY, security, type safety, pydocs, imports, logging, and test coverage
Run precommit.sh before committing — invoke after any code or documentation edit, and before staging. It modifies files in place, so re-check the diff afterward. Applies isort, black, ruff, mypy, codespell, and markdownlint in sequence, stopping at the first…
Add a new agent skill end-to-end — invoke when creating a new file under `.agents/skills/`. Runs the composition gate, chooses the name and `ref-` prefix, drafts a triggering description and a body that satisfies the effectiveness and structural standards,…
Review an existing agent skill (`.agents/skills/**/SKILL.md`) or the project `AGENTS.md` — invoke when editing or auditing a skill or `AGENTS.md`; also dispatched by `review-changes` on changes to those paths. Runs the composition, effectiveness, structural,…
Run the full unit test suite with coverage — invoke to verify a change across the whole project; use tests-run-file when assessing one source file's per-file coverage. Reports test failures and uncovered lines
Run a single test file's tests with coverage — required for assessing per-file coverage of a single source file