ponytail
Enforce the smallest working scope with existing code, standard libraries, and native features. Invoke only when the user asks for a minimal or simplest solution, YAGNI, less code, or relief from over-engineering, bloat, boilerplate, or unnecessary dependencies. Do not invoke merely because the user mentions ponytail, names a skill, or writes 'skill <word>'.
- License 许可
- MIT
- 最近更新
- 2026-11-12
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- ponytail
- license
- MIT
- description
- Enforce the smallest working scope with existing code, standard libraries, and native features. Invoke only when the user asks for a minimal or simplest solution, YAGNI, less code, or relief from over-engineering, bloat, boilerplate, or unnecessary dependencies. Do not invoke merely because the user mentions ponytail, names a skill, or writes 'skill <word>'.
- disable-model-invocation
- true
03 SKILL.md 正文
Ponytail
Runtime mode
The ponytail extension owns the active mode and persists it with the current
session branch. Default: full.
- Commands:
/ponytail off|lite|full|ultra. /ponytailand/ponytail statusreport the active mode.- The exact input
stop ponytailornormal modesets the mode tooff. - Resuming a branch restores that branch's newest mode entry.
- Child subagents start at the default mode.
Invoking $skill:ponytail adds context and does not change the runtime mode.
Selection ladder
Select the first rung that applies:
- The change has no present requirement: skip it and report
skipped: <change>, add when <trigger>. - The codebase already provides a helper, util, type, or pattern: reuse it.
- The standard library provides it: use it.
- A native platform feature covers it: use it.
- An already-installed dependency solves it: use it. Add no new dependency for a change a few lines can make.
- It can be one line: write one line.
- Otherwise: write the minimum code that works.
Run the ladder only after reading the task, reading every file the change touches, and tracing the flow end to end. When two rungs apply, take the higher rung and continue.
Rules
- Add no abstraction with one implementation, no factory with one product, and no configuration for a value that does not change.
- Add no boilerplate and no scaffolding for later.
- Delete before adding.
- Use the fewest files and the shortest diff that works.
- When two standard-library options are equal in size, choose the one that is correct on edge cases.
- For a complex request: ship the minimum version and state the unbuilt part in the same response.
- A deliberate shortcut can carry one intent comment naming its ceiling and upgrade trigger. Do not add branded prefixes to comments.
Bug fixes
- A report names a symptom.
- Grep every caller of each function the fix touches.
- Fix the root cause at the shared function all callers route through.
- A fix that patches only the reported path leaves sibling callers broken.
Intensity
| Level | Change |
|---|---|
lite |
Build what was asked and name the lazier alternative in one line. |
full |
Enforce the ladder. Standard library and native first. Shortest diff. |
ultra |
Delete before adding and challenge every requirement with no present need. |
Output
- Code first.
- Then at most three lines: what was skipped and the trigger to add it.
- Pattern:
skipped: <X>, add when <Y>. - Add no unrequested prose. Explanation the user explicitly requested is not unrequested prose.
Exclusions
Preserve input validation at trust boundaries, error handling that prevents data loss, security measures, accessibility features, and anything the user explicitly requested. When the user requires the full version, build it without re-arguing.
Verification
Non-trivial logic (a branch, a loop, a parser, a money path, or a security path) leaves exactly one runnable check: the smallest check that fails on a logic error. Trivial one-liners need no check.
Hardware
For hardware, keep one calibration parameter per physical quantity that drifts or reads off. A minimal model cannot encode physical variation.
Boundaries
Ponytail governs what is built, not prose style. The extension, not this skill, controls the runtime mode.
04 同仓库更多 Skills
writing-for-agents
Reference for writing documents an agent consumes: skills, AGENTS.md or CLAUDE.md, and docs reached by a pointer. Invoke when creating or editing a skill or an agent instruction file, or when a document's pointer wording, load, or structure is in question. Do not invoke merely because the user mentions writing-for-agents, names a skill, or writes 'skill <word>'.
wizard
Generate an interactive bash wizard for steps only a human can perform. Invoke when the user asks for a script or wizard for manual setup: infrastructure provisioning, credentials or CI secrets, a third-party dashboard, or a one-off migration or cutover. Do not invoke for steps the agent can perform, or merely because the user mentions wizard, names a skill, or writes 'skill <word>'.
wait-what
Stop and re-pitch the last message in plain, simplified language.
to-tickets
Split an agreed specification into reviewable vertical-slice issues with explicit blockers, then publish them after confirmation.
to-spec
Turn the current conversation into a specification and publish it after confirmation.
to-pr
Create a pull request from the current branch.