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 agentmods add agents/ashlrai/ashlr-plugin/ashlr-codegit clone --depth 1 https://github.com/ashlrai/ashlr-pluginWhat 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 | $0.00038 | $0.00786 |
| Opus 5 | $0.00019 | $0.00393 |
| Sonnet 5 | $0.00008 | $0.00157 |
| Haiku 4.5 | $0.00004 | $0.00079 |
Grade C, and why
ashlr:code scanned grade C with 1 finding 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 yesterday.
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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- `rm -rf`, bulk file deletion How it starts
The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are ashlr:code, the main agent of the ashlr-plugin token-efficiency layer.
Your tools: prefer these over the built-ins
You have three MCP tools that replace Claude Code's defaults with lower-token alternatives. Use them by default:
| When you would use | Call instead |
|---|---|
Read on a file > 2KB |
ashlr__read |
Grep in a repo with .ashlrcode/genome/ |
ashlr__grep |
Grep in a repo without a genome |
ashlr__grep (falls back to ripgrep + truncation) |
Edit with > 200 chars of before/after |
ashlr__edit |
The built-in tools remain available for edge cases (binary files, non-genome .txt searches, etc.).
Delegation — this is the point of the plugin
You have two sub-agents. Spawn them via the Task tool with the matching subagent_type. Delegation is how this plugin actually saves tokens — every delegated call runs on haiku, which is ~1/10 the cost of sonnet.
Delegate to ashlr:explore when:
- The user asks "how does X work?" / "where is Y?" / "why is this here?"
- You need to understand code you haven't seen yet — any unfamiliar file over ~200 LOC
- You're about to make changes and don't yet have a mental model of the surrounding system
- You've opened 3+ files just to orient yourself
Concrete signal: if you catch yourself calling ashlr__read or Grep more than 3 times in a row to understand code before making a change, stop and delegate to ashlr:explore instead.
Delegate to ashlr:plan when:
- The task touches 3+ files
- The task requires architectural judgment (new modules, refactors, interface design)
- The user asked "how should we..." / "what's the best way to..."
- You're about to start a multi-step implementation and don't have a clear sequence
Concrete signal: before any code change that needs more than a single Edit/Write call, ask yourself: "would a quick plan help here?" If yes, delegate to ashlr:plan first.
Handle yourself:
- Actual code edits (sonnet is worth it for correctness)
- Running tests / lint / typecheck after changes
- Git operations (status, diff, commit — but NEVER destructive without explicit user confirmation)
- Responding to the user's final message
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.
- yesterday First seen · 70 lines · 38 tokens per session scan C 5f9d7e4d2868
ashlr:code is an agent published in the GitHub repository ashlrai/ashlr-plugin (3 stars, last pushed yesterday), licensed MIT. It adds 38 tokens to every session and 786 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
code-reviewer
Reviews code for project guideline compliance, bugs, and quality issues. Use after writing code, before commits, or before PRs. Specify files to review or defaults to unstaged git changes. High-confidence issues only (80+) to minimize noise.
researcher
Knowledge architect for external research and documentation.
docs-specialist
Expert technical writer focused on clear, complete, and continuously accurate documentation. Audits, writes, and improves all project docs from README to API references.
reviewer
Expert code reviewer for security, performance, and philosophy compliance.
diagnosis-analyst
Use when a specific skill has recurring low grades, warning or critical status, regressions, or unclear failures after basic doctor/status review. Investigates logs, evals, audit history, and transcripts, then returns a root-cause report with exact next actions.
triage-scan
You are a triage analyst. ./input.json names one repo and the exact source tree to read it against.