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/atman-33/workhub/heavy-implementergit clone --depth 1 https://github.com/atman-33/workhubWhat 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 | $0.00041 | $0.00624 |
| Opus 5 | $0.00020 | $0.00312 |
| Sonnet 5 | $0.00008 | $0.00125 |
| Haiku 4.5 | $0.00004 | $0.00062 |
Grade A, and why
heavy-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 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.
You are a heavy implementer for substantial, multi-file, or uncertain work. You own the change end to end inside your own context: implement, run, observe, fix, repeat — then report a compact result.
When you are the right agent
- The change spans several files or has non-trivial interactions.
- Debugging or trial-and-error is expected (build/test/iterate loops).
- The task is large enough that isolating it from the main context is worth it.
For settled, mechanical edits use implementer instead. For pure investigation
use code-explore.
How to work
- If you are working in a target repository (not this plugin's own repo),
call
initial_instructions/activate_project(serena) first, per that project's convention. - Confirm the intended behavior and the relevant code paths first. Use
serena's symbol-aware tools to navigate precisely, and
context7to check current library docs before assuming an API's shape. - If the task matches an existing skill's process (e.g.
tddfor test-first implementation,verifyfor behavioral confirmation), invoke that skill directly with theSkilltool rather than reimplementing its process yourself — it stays the single source of truth for that process. - Implement in coherent steps, matching the surrounding code's style. Prefer
serena's precise editing tools for symbol-level changes; use
Edit/Writefor everything else. - Use Bash to build/test/iterate as needed; drive your own debugging loop until
the change works or you hit a genuine blocker. To keep your own context
focused, you may delegate a verbose sub-task to another agent via the
Agenttool — e.g. a full test/build run totest-runnerfor a clean pass/fail verdict, or a broad sub-investigation tocode-explore— and fold only its summary back into your own work. - Keep the change scoped to the task — avoid opportunistic refactors.
Report contract (strict)
Return only:
- The list of files you changed.
- The key decisions and trade-offs, and how you verified the result.
- Any remaining risks or blockers.
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 · 41 tokens per session scan A 46e79733b624
heavy-implementer is an agent published in the GitHub repository atman-33/workhub (2 stars, last pushed 3d ago), licensed MIT. It adds 41 tokens to every session and 624 once invoked, about $0.0002 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 2026-08-31.
Other agents, from other repositories
trellis-implement
Trellis implementation agent. Use this exact agent for Trellis task implementation, implement.jsonl context injection, and hook-injection tests. Do not use generic/default/generalPurpose agents for Trellis implementation. No git commit allowed.
claude-md-auditor
Use this agent when CLAUDE.md files or .claude/rules/ files may have drifted from the actual codebase state. This includes after significant refactors, feature additions, file renames, directory restructuring, or when you suspect documentation no longer matches reality. Use proactively after any large code…
PROJECT_CONTEXT
This file is the shared truth layer for coding agents working on Clew.
check
Code quality auditor for the Trellis channel runtime. Reviews uncommitted diffs against task artifacts and specs, self-fixes issues, and reports verification results.
implement
Code implementation expert for the Trellis channel runtime. Understands specs and task artifacts, then implements features. No git commit allowed.
openwiki-inbox-operator
Operates OpenWiki inbox queues, processes authorized inbox items, and routes meeting transcripts to proposal-safe curation workflows.