Ponytail is a pure prompt skill that remains active for the entire session once enabled: before making changes, list all callers, tests, and configurations that must be touched; avoid adding unnecessary abstractions, options, or dependencies; when fixing bugs, first grep all callers and then fix the root cause in shared code; leave a small test or assertion for any new non-trivial logic. At the end of your response, use one or two lines to state what was skipped and what risks remain. There are three tiers: lite, full, and ultra; ultra will first question the requirements themselves.
Why it matters Choose solutions in the order of "do nothing, reuse existing code, standard library, installed dependencies, one-liner," but require updating all callers and tests; validation, security, and accessibility must not be omitted.
skills.sh installs 134297 (+1488 / 24h)
Security Audit: Not audited
- Not yet scanned
Rationale
- skills.sh had 1488 installs in the last 24 hours, with a cumulative total of 134297 (representing only attention level); MIT license
- Read SKILL.md from the repository: it contains only writing and trade-off rules, with no steps for executing commands or connecting to the network
Usage
- After installation, tell the agent "ponytail" or "use the simplest solution," switch tiers using /ponytail lite|full|ultra, and say "stop ponytail" to exit
Risks
- Has not yet undergone automated desktop auditing; the website does not display installation commands
- The same repository includes companion skills such as review, audit, and debt; installing only this one changes only the coding approach