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 skills/edimuj/tokenlean/debug-bugnpx skills add edimuj/tokenlean --skill debug-buggit clone --depth 1 https://github.com/edimuj/tokenleanWhat 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 | $0.00057 | $0.00623 |
| Opus 5 | $0.00028 | $0.00311 |
| Sonnet 5 | $0.00011 | $0.00125 |
| Haiku 4.5 | $0.00006 | $0.00062 |
Grade A, and why
debug-bug 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 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.
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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Debug Bug
Systematically find and fix bugs while reading as little code as possible.
Workflow
Reproduce → Localize → Trace → Fix → Verify
1. Reproduce
tl run "<repro command>" # Capture the actual error output
No fix without reproduction. If no repro command is given, ask for one. If the bug is in tests, run the failing test. This step is non-negotiable.
2. Localize
Narrow down where the bug lives:
tl parallel \
"errors=tl errors" \
"blame=tl blame <file>" \
"history=tl history <file>"
If the error includes a stack trace, start with the top frame. If it's vague, use tl search "<error message>" to find where the error originates.
3. Trace
Understand the code path without reading entire files:
tl parallel \
"flow=tl flow <function> <file>" \
"deps=tl deps <file>" \
"snippet=tl snippet <function> <file>"
4. Fix
Apply the minimal fix. Then check for side effects:
tl parallel "guard=tl guard" "impact=tl impact <file>"
5. Verify
tl parallel "repro=tl run '<repro command>'" "tests=tl run '<test command>'"
Decision tree
Bug report → Can you reproduce it?
├─ Yes → What layer does the error come from?
│ ├─ Stack trace points to a file → tl blame + tl snippet on that file
│ ├─ Error is vague / no stack → tl errors + tl search for the error message
│ └─ Happening in dependency → tl deps to confirm, check dep version
├─ No → Ask for reproduction steps before proceeding
└─ Intermittent → tl history on suspect files, look for recent changes
Tips
- Use
tl parallelto run context-gathering commands simultaneously - Use
tl search "<error message>"to find where errors are thrown - Check
tl difffor recent regressions — the bug may be in the last few commits - If a file is under 150 lines, just read it directly — tokenlean overhead isn't worth it
- After fixing, run
tl impactto ensure dependents aren't affected by the change
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 · 82 lines · 57 tokens per session scan A 1e0a4f956016
debug-bug is a skill published in the GitHub repository edimuj/tokenlean (11 stars, last pushed 1mo ago), licensed MIT. It adds 57 tokens to every session and 623 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
localization-toolkit
This skill should be used when setting up, auditing, or enforcing internationalization/localization in UI codebases (React/TS, i18next or similar, JSON locales), including installing/configuring the i18n framework, replacing hard-coded strings, ensuring en-US/zh-CN coverage, mapping error codes to localized messages…
vs-alias-mapping
Provide system alias mapping for Search CLI. Invoke this skill when user mentions "Search CLI", "searchcli", or tries to execute searchcli commands.
jsonl-subprocess-runner
JSONL subprocess management patterns for Untether engine runners. Covers JsonlSubprocessRunner base class, template methods, event translation, session locking, resume tokens, process lifecycle, PTY usage, and concurrent session registries.
codex-opencode-pi
Codex CLI, OpenCode CLI, and Pi CLI runner protocols for Untether. Covers JSONL event types, event translation, resume mechanisms, and key differences between engines. All three are non-interactive (no control channel).
ontology-mapper
Map materials science terms, crystal structures, and sample descriptions to standardized ontology classes and properties — resolve natural-language concepts to ontology entries with confidence scores, translate Bravais lattice types, space groups, and lattice constants into ontology-compliant annotations, and produce…
localization
Localize (not just translate) text, UI content, or marketing copy to feel native in the target culture.