STOP and confirm with user if next action will push session over $10
Never retry a failed approach more than once without diagnosing root cause
Never search with Agent tool if Grep/Glob can find it in 1-2 queries
Never read entire files when offset/limit suffices
$0 cost operations (file reads, git log) are always preferred over API calls
Guard Rail
Do not initiate any action estimated to cost > $2 without confirming total session spend first.
1---2name: cost-discipline3description: Cost Discipline4---56# Cost Discipline78## Role9Own session economics as measurable constraint enforcement, not aspirational frugality.1011## Working Mode12Map -> Separate evidence from hypothesis -> Smallest intervention -> Validate.1314## Focus Areas151. Session budget cap -- $10/session hard limit, STOP and confirm before exceeding162. Token efficiency -- prefer targeted reads over full file dumps173. Batch operations -- group related API calls, never loop one-at-a-time184. Retry discipline -- max 1 retry per approach, diagnose before retrying195. Search efficiency -- targeted grep/glob before broad Agent searches206. Output verbosity -- concise responses; skip preamble and trailing summaries217. Tool selection -- use dedicated tools (Grep, Glob, Read) not Bash grep/find228. Spend tracking -- report cumulative cost at natural session milestones2324## Quality Gates25- verify: Session cost stays under $10 (track via token counts or CC cost meter)26- confirm: Each retry is preceded by diagnosis, not blind repetition27- check: Batch ops used when 3+ similar items need processing28- ensure: No full file reads when only a section is needed (use offset/limit)29- call_out: Flag if approaching $8 threshold (80% of cap)3031## Output Format32Single status line at session milestones:33```34Cost check: $X.XX / $10.00 (XX%) — [current_task] — [items_remaining]35```3637## Constraints38- STOP and confirm with user if next action will push session over $1039- Never retry a failed approach more than once without diagnosing root cause40- Never search with Agent tool if Grep/Glob can find it in 1-2 queries41- Never read entire files when offset/limit suffices42- $0 cost operations (file reads, git log) are always preferred over API calls4344## Guard Rail45Do not initiate any action estimated to cost > $2 without confirming total session spend first.
Run npx skillmds@latest add breverdbidder/cost-discipline in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Cost Discipline It is listed under Coding & Dev Tools on SkillMD.
This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
breverdbidder (@breverdbidder) published this skill. Their other Agent Skills are listed on their SkillMD profile.