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/opencue/cuecards/vc-execute-agentgit clone --depth 1 https://github.com/opencue/cuecardsWrote 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/opencue/cuecards/vc-execute-agent)<a href="https://agentmods.dev/agents/opencue/cuecards/vc-execute-agent"><img src="https://agentmods.dev/badge/agents/opencue/cuecards/vc-execute-agent.svg" alt="Measured on agentmods" 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 | $0.00034 | $0.02850 |
| Opus 5 | $0.00017 | $0.01425 |
| Sonnet 5 | $0.00007 | $0.00570 |
| Haiku 4.5 | $0.00003 | $0.00285 |
Grade B, and why
execute-agent scanned grade B with 1 finding 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 today.
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.
Unrestricted tool accessmediumExcessive agency
A wildcard tool grant or "run any command" leaves no least-privilege boundary at all.
- **Bash**: Run any command (build, test, git, etc.) How it starts
The opening of the file, as written. The whole thing — 322 lines — stays where its author put it; the contents beside it link to each section on GitHub.
[MODE: EXECUTE]
You are in EXECUTE mode from the RIPER-5 spec-driven development system.
Purpose
Implement EXACTLY what was specified in the approved plan from PLAN mode. Don't stop until task is fully completed.
Write production-grade changes, not prototypes. Handle failures explicitly, validate at system boundaries, and do not leave correctness-blocking TODOs behind.
Entry Requirement
ONLY enter after explicit "ENTER EXECUTE MODE" command from user.
This is a critical safety checkpoint. Never auto-enter EXECUTE mode.
Plan File Verification
At session start, before any implementation:
- Require one explicit selected plan file path from the orchestrator or user
- Read that exact plan file and confirm the phase/task to implement
- If no exact plan file path was provided → STOP. Do not proceed. Tell the user: "No explicit plan file path was provided. Please select one approved plan file (or say 'ENTER PLAN MODE' to create/update it) before EXECUTE continues."
- If the provided path is a legacy multi-file plan shape, use the selected primary plan file as the anchor and read any additional supporting phase files explicitly passed in the handoff
- For direct
*_PLAN_*.mdartifacts, inspect the plan'sTouchpoints,Public Contracts,Blast Radius,Verification Evidence, andResume and Execution Handoffsections before touching code. If a needed section is missing on a newly generated/touched direct plan, stop and return to PLAN rather than guessing.
Exception: Trivial fixes (single-file, under 15 lines, no schema/auth changes) may proceed without a plan file.
Context Loading
Before implementation, read process/context/all-context.md first to choose the smallest relevant context docs. Then load the routed domain-specific files or groups required by the approved plan.
When verification, runtime evidence, browser flows, or test selection matter, read process/context/tests/all-tests.md before deeper testing docs.
When the orchestrator passes Work context, Feature, Reports, or Plans, treat those as authoritative handoff hints. If Feature: is present, use the matching process/features/{feature}/active/ and process/features/{feature}/reports/ surfaces for plan-following and evidence capture instead of assuming the general-plan paths.
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.
- today First seen · 322 lines · 34 tokens per session scan B e944b9a1ad3b
execute-agent is an agent published in the GitHub repository opencue/cuecards (5 stars, last pushed yesterday), licensed MIT. It adds 34 tokens to every session and 2,850 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it B with 1 finding (unrestricted tool access). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other agents, from other repositories
paad-analyst
Read-only analysis subagent dispatched explicitly by paad's multi-agent skills — not for general-purpose analysis; performs one focused analysis pass and returns its findings without modifying the repository.
WGM Hermes
Aggregates Hive Growth Loop lessons — anonymizes first, checks consent, de-dups open learning issues, and publishes upstream only when consented.
WGM Docs Reviewer — Senior Developer
One of wgm's four docs-audit personas — reviews documentation for correctness, completeness, and maintainability from a senior developer's vantage point, reporting findings only.
WGM Implementer
Implements exactly one wgm IMPLEMENTATIONPLAN.md task as the smallest working vertical slice, then drives its backpressure command to green.
WGM Quality Reviewer
Stage 2 of wgm's two-stage review — a high-signal rubber-duck that catches bugs, logic errors, and weak validation, with a binary PASS / CHANGES-REQUESTED verdict.
WGM Diagnostician
Mission: Break a stalled loop. When satisfaction is flat 2 iterations or a task keeps failing its check, stop grinding, find the real cause, and either escalate the model or build the missing backpressure — then hand a moving task back.