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/frankxai/Starlight-Intelligence-SystemWrote 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/commands/frankxai/starlight-intelligence-system/perf-difficult-conversation)<a href="https://agentmods.dev/commands/frankxai/starlight-intelligence-system/perf-difficult-conversation"><img src="https://agentmods.dev/badge/commands/frankxai/starlight-intelligence-system/perf-difficult-conversation/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/commands/frankxai/starlight-intelligence-system/perf-difficult-conversation"><img src="https://agentmods.dev/badge/commands/frankxai/starlight-intelligence-system/perf-difficult-conversation.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 | $0.04665 |
| Opus 5 | $0.00046 | $0.02332 |
| Sonnet 5 | $0.00019 | $0.00933 |
| Haiku 4.5 | $0.00009 | $0.00466 |
Grade A, and why
perf-difficult-conversation 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 9d 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 — 317 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/perf-difficult-conversation
Load SIP.md, VOICES.md, agents/starlight-performance.md, skills/people-intelligence/feedback-conversations.md, and the manager's Genius Profile if available. Produce a Difficult Conversation Rehearsal structured around the three-conversations framework. Save to people-intelligence/performance/difficult-<topic-slug>-<YYYY-MM-DD>.md.
Disclaimer (non-waivable, heavy)
This is rehearsal architecture for a high-stakes conversation. It is NOT legal counsel. For terminations and demotions:
- Documentation gets reviewed by employment counsel in your jurisdiction BEFORE delivery. No exceptions. Generic language ("attitude problem", "not a culture fit") is legal exposure. Specific, behaviorally-grounded, dated documentation is defensible.
- Severance, final pay, benefits continuation, non-compete enforcement, and final paperwork are jurisdictional and contract-specific. Employment counsel handles those — not this rehearsal.
- If there is any concern about retaliation, discrimination (protected class), whistleblower content, harassment claims, or pending legal action, halt the rehearsal and route the entire matter to legal + HR before continuing.
- For mass layoffs or RIFs (reductions in force), WARN Act and equivalent jurisdictional notification requirements apply — this rehearsal does NOT cover those mechanics; counsel does.
- If the recipient may be in clinical-level distress (suicidality, severe mental health crisis, recent loss), the conversation may need clinical support context before, during, and after. Coordinate with EAP / HR / clinical resources.
This is a tool for the manager who must walk into the conversation. It is one input. The lawyer, HR, and EAP are the other inputs. Use them.
Input
$ARGUMENTS
When this command fires
- Termination conversation (involuntary, for performance or other cause)
- Demotion conversation
- Hard feedback that the recipient may experience as identity-attacking even with SBI grammar
- Conflict surfacing — naming a conflict that has been operating underneath but has not been spoken
- Any conversation where the manager believes the relationship may not survive
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.
- 9d ago First seen · 317 lines · 93 tokens per session scan A c6ef105ad903
perf-difficult-conversation is a command published in the GitHub repository frankxai/Starlight-Intelligence-System (8 stars, last pushed yesterday), licensed MIT. It adds 93 tokens to every session and 4,665 once invoked, about $0.0005 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 commands, from other repositories
agent
Create and manage custom AI agents.
memories
View and manage learned memories.
mpm-session-resume
Load context from paused session.
learn
End-of-session learning loop — analyze signals, approve instincts, capture directives, refresh MEMORY.md. Invokes dreamteam learn; one source of behavioral truth shared with the team.md SESSION LEARNING step.
forget
Delete specific memories.
learn
Add new learning to memory.