grepai is a local command-line tool that searches source code by meaning and traces function call graphs, rather than matching only exact words. Developers and coding agents use it to find relevant code, understand relationships between functions, and provide focused context for changes while keeping the code on the local machine. The catalogue entries expose its commands, skills, agent, and instruction for coding-agent workflows.
Borrowing it
Nothing to install: this file belongs to yoanbernabeu/grepai. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/yoanbernabeu/grepai/main/.claude/skills/grepai/SKILL.mdgit clone --depth 1 https://github.com/yoanbernabeu/grepaiWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/yoanbernabeu/grepai/grepai)<a href="https://agentmods.dev/skills/yoanbernabeu/grepai/grepai"><img src="https://agentmods.dev/badge/skills/yoanbernabeu/grepai/grepai.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00052 | $0.01020 |
| Opus 5 | $0.00026 | $0.00510 |
| Sonnet 5 | $0.00010 | $0.00204 |
| Haiku 4.5 | $0.00005 | $0.00102 |
Grade A, and why
grepai scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 117 lines — stays where its author put it; the contents beside it link to each section on GitHub.
grepai: ranked semantic search (not a Grep replacement)
grepai search answers a natural-language query with ~10 scored chunks in one
call — a ranked starting point at a fraction of the tokens of dumping raw grep
output (--json --compact alone saves ~80%). It is a ranking layer, not an
exhaustive one: a vector top-10 can miss a relevant file that keyword grep
finds trivially. The rules below save tokens without losing recall.
Tool Choice
| Query | Tool |
|---|---|
| Exact identifiers, imports, string literals | built-in Grep / git grep — fastest, exhaustive |
Intent with a canonical syntax anchor (@main, func main(, class AppDelegate) |
Grep the anchor — many "intent" questions are exact-match queries in disguise |
| Intent with no obvious anchor ("where are errors handled?") | recall-safe combo below |
| Function relationships (callers/callees) | grepai trace — grep has no equivalent |
File patterns (**/*.go) |
Glob |
Recall-Safe Combo (cheap AND exhaustive)
grep's token cost is in dumping content lines; its recall is nearly free when you ask for file names only. For an intent query, run both cheap layers:
# 1. Ranking: ~10 scored chunks, one call
grepai search "where errors are handled and logged" --json --compact
# 2. Recall: exhaustive candidate checklist — file NAMES only, ~zero tokens
git grep -ilE 'error|handl|logg' | head -50
Read grepai's top hits first, then scan the checklist for relevant-looking files grepai did not rank — read those too. Never dump full grep content output for an intent query; the file list gives you grep's recall at ~1% of the tokens.
If grepai's top hits are docs/reports instead of code: scope with
grepai search "<query>" --path <srcdir>, or add generated content to a
.grepaiignore.
How to Use This Skill
Semantic Search
Use grepai search to find code by describing what it does:
# Search with natural language (ALWAYS use English for best results)
grepai search "user authentication flow"
grepai search "error handling middleware"
grepai search "database connection pooling"
grepai search "API request validation"
# JSON output for AI agents (--compact saves ~80% tokens)
grepai search "authentication flow" --json --compact
# Limit results
grepai search "error handling" -n 5
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago Changed · +5 lines · +9 tokens per session 63a1fc417ba4
- 7d ago First seen · 112 lines · 43 tokens per session scan A d82a72e3f324
grepai is a skill published in the GitHub repository yoanbernabeu/grepai (1,838 stars, last pushed 5d ago), licensed MIT. It adds 52 tokens to every session and 1,020 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
Sverklo Code Intelligence
Gives you deep codebase understanding — semantic search, blast-radius analysis, PR review, and health audits.
use-ivygrep
Gather focused local repository context with ivygrep before implementing, debugging, reviewing, or explaining code. Use for coding tasks involving unfamiliar paths, branch changes, stack traces, architectural relationships, callers, dependents, tests, configuration, or documentation. Prefer one bounded context pack…
systematic-debugging
4-phase debugging protocol for diagnosing and fixing failing tests.
printing-press-amend
Amend a published CLI from one of two input sources: (1) dogfood mode mines the active Claude Code session transcript for friction (missing flags, hand- rolled API payloads, silent-null returns); (2) direct-input mode accepts user-supplied asks (rename a command, add commands or feeds, fix a named bug, optionally…
systemic-issue-triage
Trigger: new issue, bug report, triage, backlog, issue flood, community report, root cause, dead-end, blocked user. Attack issues by root class, never one-by-one; fixes must shrink the system, not grow it.
report
Investigate bugs comprehensively — cascade through trace, capture browser evidence, extract observability data, and prepare or explicitly create a GitHub issue with grounded findings.