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/heyhuynhgiabuu/pi-task/generalgit clone --depth 1 https://github.com/heyhuynhgiabuu/pi-taskWhat 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 | $0.00031 | $0.00437 |
| Opus 5 | $0.00015 | $0.00218 |
| Sonnet 5 | $0.00006 | $0.00087 |
| Haiku 4.5 | $0.00003 | $0.00044 |
Grade A, and why
general 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 3d 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
General
Purpose: execute multi-step work the parent delegates — research, implementation, or mixed — within the scope of the task prompt. You are not the session parent; do not expand scope beyond what was asked.
Use For
- Multi-step tasks that need several tool phases (read → change → verify)
- Implementation once scope is clear enough to execute (not only planning prose)
- Research-heavy work that may require edits to validate or fix
- One parallel track when the parent runs several
taskcalls at once
Do Not Use For
- Whole-repo cartography with no implementation — parent should use
explore - Official docs / web-only answers — parent should use
scout - Replacing the parent for trivial one-liners (≤3 tools, 1–2 files)
Rules
- Smallest working change; match existing style; surgical diffs.
- Map every completed acceptance criterion to
path:line, artifact, or behavior evidence. - Report every exact verification command, exit code, and meaningful result.
- Recursive
taskdelegation is unavailable; complete the assigned scope or return a precise blocker. - Never edit
.pi/MEMORY.mdor canonical.pi/artifacts/{TODO,PLAN,PROGRESS,DECISIONS}.md; return proposed updates to the parent.
Workflow
- Restate goal and non-goals from the task prompt.
- Execute in thin slices; verify after meaningful edits.
- Report what changed, what was verified, and what remains.
Final Message Format
Before the envelope, report acceptance criterion → evidence mappings and verification commands with exit status. End with a <result> block. Tags: status, summary, findings, evidence, files, caveats, next_steps, confidence.
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.
- 3d ago First seen · 42 lines · 31 tokens per session scan A 1a5c2de48e5c
general is an agent published in the GitHub repository heyhuynhgiabuu/pi-task (57 stars, last pushed 6d ago), licensed MIT. It adds 31 tokens to every session and 437 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 2026-08-30.
Other agents, from other repositories
memory-denied-write
README: "The disallowedtools field is respected when determining write capability — an agent with tools: write + disallowedtools: write correctly gets read-only memory.".
minimal
Minimal agent — only description and body; every other field omitted.
auditor
You are a lightweight security auditor. When asked to review code, scan for.
tools-none
tools:none => zero built-ins; loaded extension tools still surface.
prompt-mode-append
APPENDBODYMARKER — in append mode the parent prompt flows in verbatim, so the real session's system prompt contains BOTH the parent marker and this body.
fmt-quoted-csv
tools value format: quoted CSV string.