Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add bnomei/frigg --skill frigg-first-code-searchgit clone --depth 1 https://github.com/bnomei/friggWrote 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/bnomei/frigg/frigg-first-code-search)<a href="https://agentmods.dev/skills/bnomei/frigg/frigg-first-code-search"><img src="https://agentmods.dev/badge/skills/bnomei/frigg/frigg-first-code-search/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bnomei/frigg/frigg-first-code-search"><img src="https://agentmods.dev/badge/skills/bnomei/frigg/frigg-first-code-search.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00080 | $0.12100 |
| Opus 5.5 | $0.00032 | $0.04840 |
| Sonnet 5.5 | $0.00016 | $0.02420 |
| Haiku 4.5 | $0.00008 | $0.01210 |
Grade A, and why
frigg-first-code-search 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 1mo 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 — 860 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Frigg First Code Search
Frigg is the default for code discovery, file listing, navigation, exact code search, and bounded source reads.
Before using shell rg, grep, find, fd, cat, or sed for code exploration, use the matching Frigg MCP tool in attached or attachable code repositories.
Shell tools are fallback only for git state and diffs, non-code files, build/test output, generated or unindexed files, explicit live-disk verification, or when Frigg is unavailable.
Do not run parallel shell grep in the same turn as Frigg search on indexed source.
Follow the evidence, not just the shell replacement: use search_symbol for a known
identifier and search_batch for independent hypotheses when callable. After selecting a hit,
copy its target_ref into navigation when relationships matter, then use read_match for source
proof. A known path still warrants direct read_file; do not add navigation merely to use more tools.
Harness boundary (Frigg is an evidence layer, not an agent OS):
- Host tool order, Grep-first defaults, and intermittent MCP registration/inheritance flukes are outside Frigg product code (FUT-003 harness boundary — not a Frigg CI gate or ranking bug).
- When the bridge is down or tools are unregistered, that is harness reliability, not ranking quality.
- Soft PreToolUse / adopt hooks are opt-in (not guaranteed Frigg-first). Soft only: Claude hook injects next-step context (
search_text/search_batch/ …) and never hard-denies Grep/shell (permissionDecisionis not product). - Hard block of Grep/
rgon indexed roots is a host experiment, not Frigg core — do not inventFRIGG_HOOK_STRICTor product tool-order hacks. - Do not invent product hacks (fake Grep, hide host tools, force MCP tool order, or a “subagent bridge health” Frigg tool — Frigg cannot see the host registry when it is not registered).
- Route indexed source through Frigg when the live tool list exposes Frigg tools; progressive disclosure on this top screen is intentional (30-second rule).
Probe-on-spawn (Task / subagent / new session):
1. Before claiming Frigg-first, verify Frigg is on the *live* host tool surface for THIS spawn:
live tools/list (or a successful Frigg CallMcpTool). Do NOT trust on-disk mcps/…/tools
descriptors or a parent session's tool list as proof this spawn has Frigg.
2. If Frigg is absent from live tools/list for this spawn (empty Frigg server, or
Tool not found when calling a known Frigg tool like search_text while the parent still has Frigg):
- Shell/Grep/Read fallback is correct for this spawn
- Note the gap once (harness MCP inheritance / registration) — do NOT file as Frigg search/ranking failure
- Do NOT invent workspace_reindex / bridge_health / fake tools to "fix" registration
- (Invented names / extended-only tools missing while other Frigg tools work → surface honesty,
not inheritance — see tool-surface rules)
3. If the initial Frigg workspace is the wrong repository, attach the task repository with
workspace(path=...) before falling back to local tools.
4. Parent session success does NOT imply child Task success — re-probe every delegation.
5. Prefer loopback HTTP + frigg serve when multiple clients/subagents should share one Frigg
*after* each spawn actually has Frigg registered; HTTP shares state when registration works —
it does not create child registration by itself.
Partial surface (some Frigg tools work, recommended ones are missing):
- Compare the host-callable tools, the installed skill's
mcp.jsonincludeTools(or inlinemcpServers), andworkspace.runtime.tools_exposed/ livetools/listwhen available. A stale skill filter can hide symbol, batch, and navigation tools even when the server supports them. - Report the reduced surface once. Do not call host-unavailable tools just because the skill or server lists them. Use parallel scoped searches when batch is unavailable; use bounded exact identifier/import/call-site searches when navigation is unavailable, then read the witnesses.
- See runtime recovery before refreshing an install. Preserve its endpoint/credentials; an expired HTTP session is not evidence of a smaller server tool surface.
What ships with it
10 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- .claude-plugin/plugin.json 633 B
- .mcp.json 111 B
- agents/openai.yaml 1.2 KB
- hooks/hooks.json 262 B
- mcp.json 534 B
- references/discovery-and-evidence.md 16 KB
- references/extended-tools.md 1.7 KB
- references/navigation-and-structure.md 11 KB
- references/workflows.md 5.3 KB
- references/workspace-and-runtime.md 11 KB
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.
- 1mo ago Changed · +28 lines d9917573e2d7
- 1mo ago Changed · +829 lines · +80 tokens per session 39d720b45709
- 1mo ago First seen · 3 lines · 0 tokens per session scan A a95e8f3dda7c
frigg-first-code-search is a skill published in the GitHub repository bnomei/frigg (5 stars, last pushed 17d ago), licensed MIT. It adds 80 tokens to every session and 12,100 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-08-31.
Other skills, from other repositories
mcpls
Install, configure, and run the mcpls CLI — a Rust binary bridging MCP to LSP that gives an agent compiler-grade code intelligence (hover, definitions, references, diagnostics, rename). Use when setting up or registering mcpls as an MCP server, writing or debugging an mcpls.toml, choosing CLI flags or MCPLS…
windows-compat
Audit and harden this Rust repo (code-graph-mcp) for Windows correctness: path-spelling drift between producers, the 32,767-char command-line cap, index-key mismatches, and path predicates that assume one ecosystem's layout. Use whenever touching code that builds, compares, prints, or stores a filesystem path; that…
index
Diagnose and fix code-graph index issues. Use when: search returns unexpected/empty results, or after major codebase restructuring. Covers the health check and both rebuild paths from the CLI.
explore
Understand code structure with the AST index before reading files one by one — when starting work in unfamiliar code, exploring a module before a change, or finding which file to edit. An overview lists a directory's symbols and their callers in one call.
scip-integrity-audit
Investigate whether a feature fulfills its promised behavior through live consumers. Find partial implementations, decorative checks, ineffective parser migrations, competing implementations, misleading results, and tests validating their own assumptions.
scip-query
Use code relationships, module structure, quality findings and change impact to plan and review changes alongside ordinary source-reading tools.