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 markmhendrickson/ateles --skill writegit clone --depth 1 https://github.com/markmhendrickson/atelesWrote 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/markmhendrickson/ateles/write)<a href="https://agentmods.dev/skills/markmhendrickson/ateles/write"><img src="https://agentmods.dev/badge/skills/markmhendrickson/ateles/write.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00014 | $0.01462 |
| Opus 5 | $0.00007 | $0.00731 |
| Sonnet 5 | $0.00003 | $0.00292 |
| Haiku 4.5 | $0.00001 | $0.00146 |
Grade A, and why
write 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 4d 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 — 118 lines — stays where its author put it; the contents beside it link to each section on GitHub.
[Body content following all content_style_enforcement rules]
Accompanied by `.summary.md`, `.tweet.md`, `.linkedin.md`, and `.share.md` files.
## Viral structure principles
These principles are derived from analyzing which posts break out (15-20x baseline) versus baseline. Apply to every draft.
### Lead with the most surprising single fact or moment
The punchline IS the first sentence. Never set up context before the hook.
### Disaster-then-redemption is the highest-performing narrative structure
If the piece contains a failure, mistake, or crisis followed by recovery or insight, lead with the disaster. Delay the resolution.
### Personal vulnerability + structural insight = breakout
"I lived this" + "here's the systemic reason why." Neither alone is enough.
### Absolve individuals, indict systems
Posts that say "not a failure of people, a failure of incentive structure" get retweeted because people share things that absolve them while explaining their experience.
### Never resolve the tension in the main tweet
Leave one thread open. The reply chain is where the strongest algorithmic signal lives.
### End with stakes or questions, never summaries
The best questions are ones the audience has personal experience with. Could 5 different readers give 5 different real answers? If not, rephrase. Rhetorical questions do not count.
### Make readers feel like insiders
Shareable one-liners make the reader feel smarter for having read them. These are the phrases that get screenshot-shared and bookmarked. Extract or compress one per piece.
### Disagree with bigger accounts, respectfully and specifically
Affirm the problem, challenge the solution. The disagreement should be structural, not personal.
## Style notes
### Community engagement voice
Reddit replies, Substack comments, GitHub discussions, and forum posts use a warmer register than social media or website posts. The goal is rapport and credibility with peer builders, not compression or virality.
**Soften authority on others' work.** Never issue definitive architectural verdicts on someone else's system. "That's the right idea" not "That's the right architecture." Acknowledge and ask, don't evaluate and pronounce. The author respects that other builders know their own system better than he does.
**Warmer punctuation.** Use "Thanks!" not "Thanks." when responding to someone who shared useful work or a substantive reply. Brief warmth signals genuine appreciation without being effusive.
**First-person question closers.** "I'm curious about the transport layer specifically" not "Curious about the transport layer." Community register is more conversational than Twitter's telegraphic compression. Adding "I'm" and specificity makes the question feel like genuine interest rather than an interview prompt.
**Sentence flow over staccato.** Prefer joining short related clauses into one flowing sentence with lighter punctuation (commas, en dashes) over statement-period-statement sequences. Community replies read as conversation, not bullet points.
**Em dashes and en dashes are acceptable** in community engagement. Same exemption as social media drafts. The `content_style_enforcement` no-dash rule applies to website posts, docs, and marketing copy only.
**End with a real question.** Every community reply should close with a specific, answerable question that demonstrates you read their work and want to understand their system better. Generic "thoughts?" closers don't count. "How does your policy layer decide what gets written?" counts.
### Em dashes in social content
The `content_style_enforcement` rule prohibits em dashes in blog posts, docs, and marketing copy. Social media drafts and community engagement are exempt. Em dashes are effective in tweets for compression and pacing.
### Pain-first vocabulary for social
Lead with pain and outcomes in hooks. Use architectural language only as explanation underneath.
| Internal (use in replies, threads, deep dives) | Social (use in hooks, main tweets, punchy takes) |
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.
- 4d ago First seen · 118 lines · 14 tokens per session scan A f40b5b6972bd
write is a skill published in the GitHub repository markmhendrickson/ateles (6 stars, last pushed today), licensed MIT. It adds 14 tokens to every session and 1,462 once invoked, about $0.0001 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 2026-09-03.
Other skills, from other repositories
local-llm-tool
Local LLM execution tool for text generation and chat through Ollama or vLLM endpoints. Use when: running on-prem inference, calling a local GPU model, or summarizing with a self-hosted LLM.
aaai
AAAI paper formatting — activate when the user wants AAAI template setup, migration, or formatting/compilation fixes.
colm
COLM (Conference on Language Modeling) paper formatting — activate when the user wants COLM template setup, migration, or formatting/compilation fixes.
review-pr
A pull-request review skill for evaluating proposed code changes and posting a quality verdict. A pull request is a request to merge changes into a shared codebase.
tasks
A task-breakdown tool that turns a plan, specification, feature brief, bug report, technical-debt item, or chore into small TASK-NNN work items. Technical debt means postponed cleanup or design work in a codebase.
conference-research
Academic venue paper discovery — activate when you need to find papers accepted at specific conferences or journals (ACL, EMNLP, NeurIPS, ICML, ICLR, MLSys, etc.), analyze venue acceptance preferences, or look up CFP requirements.