RiyanKothari
- 6 skills
- 0 followers
- 16 hours ago last updated
- ▌ Ralph Loop · riyankothariRun a bounded self-correcting loop that re-feeds the same prompt every time Claude stops, until a completion promise is genuinely true or an iteration cap is reached. Use when a task has a machine-checkable finish line such as a passing test suite or a clean linter, and iterating unattended beats steering each step by hand.
- ▌ Token Harness · riyankothari bundleCut Claude Code token cost through model-tier routing, budget-capped memory recall, and evidence-gated self-scoring. Use when API cost matters, when deciding whether to delegate work to a cheaper model, when you need memory that survives across sessions without bloating context, or when auditing what is eating the context window.
- ▌ Safe Operations · riyankothariRules for actions that can destroy work — writing config files, deleting, overwriting, force-pushing, running installers. Use before any operation that modifies files you did not create, before writing to a config another tool owns, before any bulk find-and-replace, and whenever an operation is hard to undo.
- ▌ Secrets Hygiene · riyankothariKeeping credentials out of files, stores and repositories — redacting at write time, scanning before committing, and handling a leak that already happened. Use before committing or publishing anything, before storing prompt or conversation text, when initialising a repo, and whenever a tool persists user input to disk.
- ▌ Debug Systematically · riyankothariFinding the actual cause instead of the plausible one — probing real state, checking derived data, and fixing root causes rather than symptoms. Use when a fix does not work, when behaviour contradicts the code, when a test fails unexpectedly, or when tempted to guess at a cause instead of observing it.
- ▌ Verify Before Claiming · riyankothariDiscipline for proving work rather than asserting it — running the thing, measuring before optimising, and reporting honestly when results are bad. Use before saying something works, before optimising anything, when reporting a result or score, and whenever tempted to describe intent as outcome.