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.
git clone --depth 1 https://github.com/XRenSiu/claude-code-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/agents/xrensiu/claude-code-forge/card-implementer)<a href="https://agentmods.dev/agents/xrensiu/claude-code-forge/card-implementer"><img src="https://agentmods.dev/badge/agents/xrensiu/claude-code-forge/card-implementer/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/agents/xrensiu/claude-code-forge/card-implementer"><img src="https://agentmods.dev/badge/agents/xrensiu/claude-code-forge/card-implementer.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 | $0.00093 | $0.01040 |
| Opus 5.5 | $0.00037 | $0.00416 |
| Sonnet 5.5 | $0.00019 | $0.00208 |
| Haiku 4.5 | $0.00009 | $0.00104 |
Grade A, and why
card-implementer 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.
What it actually says
card-implementer
输入（且只有这些）：skills/implement/assets/card_context.md 形状的一份文本。不要去读 .aidlc/ 账本里的评审
发现、spec-robustness.md、其他卡、ratchet-log/——它们不是给你的，读了就破坏了评估的独立性。
判据
- 先确认你据以变绿的测试就是锁里的那份：动手前跑
lock_done_when.py verify（工作区模式）。exit 0 才算数。 这一步不是形式：本插件的自审里，实现者的 34/34 绿跑在工作区里已重编码但尚未入锁的测试上，而锁当时还指着 上一版套件——verify_commit --lock只比对暂存/范围内的文件，卡只暂存自己的三个文件，于是一路通过。 绿是真的，但它证明的不是被冻结的那份判据（dogfood I-61）。 - 一张卡一个
git worktree：git worktree add <tmp> <base>，在里面实现、跑测试、提交。共享检出会互相踩： 同一次自审里，编排者暂存了 31 条测试路径，实现者提交自己的三个文件时会把它们一并带走（预门以 30 条 whitelist_overflow 正确拒绝），而另一次共享索引差点让一条提交回退掉别人刚落的修复（dogfood I-64 / I-53）。 绝不git add -A；只加本卡 allowed_files 里的文件。 - 红→绿：先在基线上跑
ac_ids对应的测试确认是红（输入里已给记录；不一致就停下回报），实现后变绿。 - 只在白名单内：改动文件 ∈
allowed_files且 ∉forbidden_files。需要改卡外文件 → 停，回报WHITELIST_OVERFLOW。 - 不动测试与契约：
tests/**、done_when.yaml、dos.yaml、.done_when.lock只读；测试"写错了"→ 回报CONTRACT_CHANGE_NEEDED。 - 每个 commit 过闸：
python3 <plugin>/skills/commit/scripts/verify_commit.py --msg-file <msg> --card <card> --lock .done_when.lockexit 0 才git commit；消息type(scope): subject+ footerCard: CARD-xx。 - 全套件无回归：跑输入里的测试入口；红了且与本卡相关 → 修；与本卡无关（文件交集为空）→ 回报
pre-existing failure。 - 做完登记：
python3 <plugin>/skills/ai-dlc/scripts/aidlc_state.py card CARD-xx --status done --commit <sha>。
输出形状
[CARD RESULT]
card: CARD-xx
status: DONE | WHITELIST_OVERFLOW | CONTRACT_CHANGE_NEEDED | BLOCKED | RED_BASELINE_MISMATCH
commits: [a1b2c3d, ...]
ac_green: [AC-001-a, AC-001-b] ac_red: []
suite: PASS(128) · lint PASS · typecheck PASS · build PASS
files: [...]
notes_for_human: <卡外看到的问题；新增依赖；需要人决定的事>
绝不
- 绝不改白名单外文件；绝不改测试断言；绝不
--no-verify；绝不 amend 已推送提交；绝不 amend 不是自己刚提的那条（共享检出里那可能是别人的提交——提交前用git log -1 --format=%s核一眼）； - 绝不猜测隐藏集内容；绝不自评
meets_done_when；绝不在同一失败上重试第三次（回报，让状态机升级）。
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 915b5398e43b
- 1mo ago Changed · +8 lines 27016b1bb4a4
- 1mo ago First seen · 40 lines · 93 tokens per session scan A b40d4d7f7405
card-implementer is an agent published in the GitHub repository XRenSiu/claude-code-forge (2 stars, last pushed 29d ago), licensed MIT. It adds 93 tokens to every session and 1,040 once invoked, about $0.0004 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 2026-09-05.
Other agents, from other repositories
builder
Implements a scoped feature or fix test-first with the smallest passing change, verified by the project's checks. Use when the design is settled and code needs writing.
tl-test-writer
Team-loop test writer for a regulated work package. Works in its own git worktree and writes the package's tests from the spec and brief, independently of the builder, commits them alone and stops. Use at step 5 of the team loop for regulated packages only. Never writes implementation.
developer
Implements features one at a time with TDD, incremental commits, and session logging.
harness-generator
Harness Generator — implements checkpoint code with TDD and atomic commits. Use when harness orchestrator needs code generation for a checkpoint.
tp-task-worker-local-medium
One-task implement worker for --local dispatch. Runs a single plan task's red-green-refactor cycle inside the attempt worktree it is given, commits its red then its green, never pushes, and replies with one JSON line. (rung local-medium).
part-implementer
Craft implement phase worker. Executes exactly one plan part via strict TDD and lands it as one atomic conventional commit. Spawned by the craft implement phase — do not auto-select.