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 harshanandak/forge --skill forge-buildgit clone --depth 1 https://github.com/harshanandak/forgeWrote 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/harshanandak/forge/forge-build)<a href="https://agentmods.dev/skills/harshanandak/forge/forge-build"><img src="https://agentmods.dev/badge/skills/harshanandak/forge/forge-build/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/harshanandak/forge/forge-build"><img src="https://agentmods.dev/badge/skills/harshanandak/forge/forge-build.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 | $0.00030 | $0.00926 |
| Opus 5.5 | $0.00012 | $0.00370 |
| Sonnet 5.5 | $0.00006 | $0.00185 |
| Haiku 4.5 | $0.00003 | $0.00093 |
Grade A, and why
forge-build 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 2d 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 — 55 lines — stays where its author put it; the contents beside it link to each section on GitHub.
forge-build
Read pull-request.md for setup and PR handoff.
Every wake runs forge pr state <n> and acts on verdict.
Worker brief
Before each call run forge issue call <slice> --role <role>; LIMIT_REACHED ends the intent: report, start nothing. Retry REVISION_CONFLICT.
Fill every field before delegating; an empty field blocks the call:
Goal: <slice id and end state; read forge issue show for the spec>
Scope: <Owns and done boundary>
Context: <Touches / Needs, adopted decision titles and checkpoint>
Acceptance: <observable pass predicates>
Verify: <named red/green check, risk band, scenario and final-head checks>
Timebox: `forge issue call`'s timeboxMinutes (at most 45 minutes); stop and report the checkpoint when timeboxMinutes expire, never later than stopBy.
Forbidden: pass named environment variables only, never the whole environment or secret values; stop at push; never merge or merge base unless asked.
Report: <head, receipts, saved output and not checked>
Standing rules: <adopted rules; layer 1 passes the memory packet>
Red, then green
- Read
forge issue resume <slice>and every Touches file before editing; later comments override the body. Readforge issue comments <slice>once before acting if a checkpoint exists ormoreis true. forge-cut risk: light 0-1, standard 2-3, heavy 4-5. - Every slice goes red then green. Name a check whose expected value is independent of the implementation. Run the new assertion on trunk first; it must fail on that assertion, not setup or unrelated errors.
forge check run --input '{"checks":["<name>"]}'. Expect exit 1 andok:falsefor FAIL (exit 2 for INCOMPLETE or NOT_EXECUTED); keep the fullresultreceipt and its id even on nonzero exit. PASS alone exits 0. Do not chain red checks with&&or discard their output on failure.- Make the smallest fix, rerun to green and keep its receipt id. Use small conventional commits.
Three proof levels, scaled to risk:
- Light, not user-visible: the independent red/green assertion plus the final-head receipt.
- User-visible, including standard slices (UI, CLI output, API response, published package): also save one scenario on the built artifact as users get it, with a pass predicate and output.
- Heavy or critical: always run that scenario on trunk and head, even without a user-visible change, and get a fresh-context review against every acceptance item through
forge-reviewbefore opening.
What ships with it
2 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.
- 2d ago First seen · 55 lines · 30 tokens per session scan A 253bcaa3e3c5
forge-build is a skill published in the GitHub repository harshanandak/forge (0 stars, last pushed yesterday), licensed MIT. It adds 30 tokens to every session and 926 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 2026-10-28.
Other skills, from other repositories
team-qa
Orchestrate the QA team through a full testing cycle — qa-lead strategy and test plan, qa-tester case writing, execution, sign-off.
scaffold-dotnet-test-project
MUST USE when an existing .NET test project was excluded from a .slnf/CI solution filter, disappeared from .sln/.slnx discovery, or lost its production ProjectReference; also for requests to set up, create, reuse, add, register, include, or repair a test project. Handles "tests pass directly but CI discovers zero"…
test-result-feedback
Close the acceptance loop for a tested change — consume the apc test TierResult, map EACH acceptance criterion to a pass/fail outcome from structured fields (never chat text), attach an evidence ref, and write it back to the task. A SUT red is a finding, never softened to green.
conductor-implement
Execute tasks from a track's implementation plan following TDD workflow.
cy-execute-task
Implement and verify an existing CompozyOS spec task, then update its tracking. Excludes review remediation.
implement-feature
Implement an approved feature plan with benefit-based placement, TDD, revision-bound evidence, and PR-ready output.