Skip to main content

usage

Show what sharp costs in tokens and what it delivered, to judge its return on investment. Reads exact usage from local session transcripts; private to this machine.

Source facts

Repository
pipeshub-ai/sharp
Last source activity
October 15, 2026 at 19:46
Detected SKILL.md language
English
Stars
4
Forks
0

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
usage
description
Show what sharp costs in tokens and what it delivered, to judge its return on investment. Reads exact usage from local session transcripts; private to this machine.
argument-hint
[days, default 30]
disable-model-invocation
true
allowed-tools
Bash(node ~/.sharp/bin/sharp.mjs *)
# /sharp:usage: cost and return Measured locally over the last 30 days (if `$ARGUMENTS` names a different number of days, re-run `node ~/.sharp/bin/sharp.mjs tokens <days>`): ```! node ~/.sharp/bin/sharp.mjs tokens 30 2>/dev/null || echo "{}" ``` Write a short report (12 lines or fewer) in plain words: 1. **What sharp cost:** the notes it added (as a share of all tokens) and the extra replies it asked for, such as change summaries (tokens and share). Add the overall overhead share. Show dollars only if prices are set; otherwise say how to set them in one line. 2. **Learning you chose:** tokens spent in `/sharp:` commands. This is time you chose to invest, not overhead. 3. **What it delivered:** irreversible actions stopped for your decision, decisions recorded and looked back on, lessons saved, past lessons brought back when relevant, and how often practised lessons were applied correctly. 4. **One line of judgment:** is the overhead small relative to what it delivered? If overhead is high and value low, suggest a concrete setting change (for example `/sharp:mode light` or `#ship`). If fewer than 10 turns were measured, say the numbers are too early to judge. Never invent numbers that aren't in the data. ## How to write - Plain words. Define "token" once (a small chunk of text the AI reads or writes; cost scales with tokens). - Never mention hooks or other internals.
View on GitHub