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 IamK77/Skill --skill forgegit clone --depth 1 https://github.com/IamK77/SkillWrote 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/iamk77/skill/forge)<a href="https://agentmods.dev/skills/iamk77/skill/forge"><img src="https://agentmods.dev/badge/skills/iamk77/skill/forge/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/iamk77/skill/forge"><img src="https://agentmods.dev/badge/skills/iamk77/skill/forge.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 | $0.00034 | $0.01003 |
| Opus 5.5 | $0.00014 | $0.00401 |
| Sonnet 5.5 | $0.00007 | $0.00201 |
| Haiku 4.5 | $0.00003 | $0.00100 |
Grade A, and why
forge scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 27d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
forge
Execute a defined research experiment with traceable inputs, configurations, runs, and outputs. Use for experiment implementation and reproducibility, not to invent the research claim.
Inputs
Versioned experimental plan; implementation; data access; environment; compute authorisation and budget.
Run record
Use checklist 0.5 or newer for this SOP. Reading or reloading the skill must not reset work. For a new task, run checklist init "${CLAUDE_SKILL_DIR}" --new --path "/absolute/project"; retain its returned ID as RUN. For existing work, use checklist resume "$RUN", then checklist show --run "$RUN". Outside Claude, replace the skill-directory variable with this directory’s actual path.
Commands below record work after it is done. Replace evidence placeholders with concise artifact locations and observations. Manual confirmations are not independent proof; a sensor pass covers only the command and inputs it observed. verify does not close a stage; advance checks that its required readings are fulfilled. Use report --run "$RUN" for the retained history.
Procedure
prepare
Match code, data, and configuration to the plan. Run a small end-to-end sanity check and verify that metrics and failure records have the intended meaning.
Record these artifacts after doing the work:
- Code, data, environment, and plan versions are recorded.
checklist check prepare inputs --run "$RUN" --evidence "<artifact and observation>" - A small end-to-end run validates data flow and metric computation.
checklist check prepare sanity --run "$RUN" --evidence "<artifact and observation>"
When this stage is fulfilled: checklist advance prepare --run "$RUN".
execute
Run within the agreed budget. Preserve seeds, commands, logs, failed runs, and deviations; checkpoint where useful. Keep exploratory adjustments separate from the committed comparison.
What ships with it
7 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.
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.
- 27d ago Changed · -79 lines · -182 tokens per session 08b157ee66d5
- 1mo ago First seen · 168 lines · 216 tokens per session scan A c79a6873aa6f
forge is a skill published in the GitHub repository IamK77/Skill (2 stars, last pushed 4d ago), licensed Apache-2.0. It adds 34 tokens to every session and 1,003 once invoked, about $0.0001 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
feature-engineering
Use when building or improving time series forecasting models and the user asks about exogenous variables, calendar features, rolling statistics, cyclical encoding, differencing, or feature scaling — or when forecast accuracy has plateaued and new features may help.
vast-experiment-ops
Use when running or reporting ML experiments on rented Vast.ai GPU instances — diagnosing strict FP32 equivalence checks that pass on CPU but fail on CUDA (cuDNN TF32), parsing JSON metrics collected over SSH, and keeping held-out/test-scope reporting accurate.
tabpfn-regress
Run a TabPFN regression baseline, generate the first submission, then optimize with GBT ensembles and regression-specific post-processing (clipping, target transforms, rank blending). Use after tabpfn-explore has prepared the data and CV folds.
saelens
Train sparse autoencoders to interpret model features.
nnsight-remote-interpretability
Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any PyTorch architecture.
transformer-lens-interpretability
Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing activation patching experiments.