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 voicelogica/voice-logica-plugin --skill fix-agent-behaviorgit clone --depth 1 https://github.com/voicelogica/voice-logica-pluginWrote 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/voicelogica/voice-logica-plugin/fix-agent-behavior)<a href="https://agentmods.dev/skills/voicelogica/voice-logica-plugin/fix-agent-behavior"><img src="https://agentmods.dev/badge/skills/voicelogica/voice-logica-plugin/fix-agent-behavior/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/voicelogica/voice-logica-plugin/fix-agent-behavior"><img src="https://agentmods.dev/badge/skills/voicelogica/voice-logica-plugin/fix-agent-behavior.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 | $0.00061 | $0.01010 |
| Opus 5 | $0.00030 | $0.00505 |
| Sonnet 5 | $0.00012 | $0.00202 |
| Haiku 4.5 | $0.00006 | $0.00101 |
Grade A, and why
fix-agent-behavior 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 10d 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 — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Fix Voice Logica agent behavior
Most "the platform is down" reports are prompt or config tweaks. Read the live agent and a Call ID before changing anything.
Start with
- Call ID
- Expected vs actual
- Language of the call
- Whether this used to work
Then decide: the agent broke a written rule, followed a rule the user now wants changed, or the prompt is silent. Those need different edits.
If the rule is already in the prompt and was violated, restating it again wastes system-prompt budget (get_subscription before adding text). Fix the mechanism (tool, transfer prompt, end-call, language) instead.
If get_agent_prompt is approaching or above ~20,000 characters, do not add more rules. Move facts to knowledge and trim first (edit-voice-agents). A 30k+ prompt needs router + specialist, not another paragraph.
If the agent speaks JSON, tool names, or brackets out loud, collect Call ID + timestamp and file a ticket (create-support-ticket). Do not try to prompt-paper over leaked system content.
Common fixes
Off-topic callers. Add an early rule: detect wrong-industry intent, say you cannot help, end the call without collecting contacts.
Too many questions. Collect only fields required for that branch. A message-only path often needs name + callback number. "Anything else?" delays end-call; if they want it gone, use End Call config for that branch.
Re-asks. Add "never re-ask if already provided in this call". Also check language mismatch and interruption / skip_turn.
Wrong language. Confirm languages enabled on the agent, put a follow-the-caller-language rule at the top of the prompt, then retest.
Greeting restarts. Deduplicate welcome message vs prompt greeting. Add a resume-after-failed-transfer rule.
Ends call too early or says goodbye then continues. End-call reasons must match real completion states. Never put "transferring" in endCallReasons. Finish message-taking before end-call.
Interrupts the caller. Do not hunt a silence-threshold knob on Soniox agents (waitForFullTranscription and silenceDurationMs are inert). skip_turn is the turn-taking lever. Knobs: the tool prompt and timeoutSeconds (1 to 10).
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.
- 10d ago First seen · 66 lines · 0 tokens per session scan A 34b19b12833a
fix-agent-behavior is a skill published in the GitHub repository voicelogica/voice-logica-plugin (0 stars, last pushed 18d ago), licensed MIT. It adds 61 tokens to every session and 1,010 once invoked, about $0.0003 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 skills, from other repositories
gws-gmail-send
Skill "gws-gmail-send" from googleworkspace/cli, covering gmail +send, usage, flags, examples and tips.
gws-shared
The gws binary must be on $PATH. See the project README for install options.
gws-sheets
Google Sheets: Read and write spreadsheets.
gws-gmail
Gmail: Send, read, and manage email.
gws-modelarmor
Google Model Armor: Filter user-generated content for safety.
gws-docs
Read and write Google Docs.