System Prompts Leaks is a collection of captured system instructions used to guide AI chatbots and coding agents before they receive user messages. It serves researchers and developers studying how different AI assistants are directed.
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 asgeirtj/system_prompts_leaks --skill facebook-cligit clone --depth 1 https://github.com/asgeirtj/system_prompts_leaksWrote 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/asgeirtj/system_prompts_leaks/facebook-cli)<a href="https://agentmods.dev/skills/asgeirtj/system_prompts_leaks/facebook-cli"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/facebook-cli/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/asgeirtj/system_prompts_leaks/facebook-cli"><img src="https://agentmods.dev/badge/skills/asgeirtj/system_prompts_leaks/facebook-cli.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 | $0.00118 | $0.05364 |
| Opus 5.5 | $0.00047 | $0.02146 |
| Sonnet 5.5 | $0.00024 | $0.01073 |
| Haiku 4.5 | $0.00012 | $0.00536 |
Grade A, and why
facebook_cli 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 12d 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 — 200 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Facebook CLI
Pasted links first
If the user pasted a Facebook Marketplace item link or a share link, never
open it with the browser: the browser cannot pass the Facebook login wall.
For an item link, run facebook-cli marketplace listing details --url '<pasted link>'. For a share link, decode it first with facebook-cli link-sharing decode-url --url '<pasted link>', then follow
references/marketplace.md Step 0 for the decoded URL.
Quick Reference
The read-only post and Marketplace-only write restrictions below apply to personal-profile activity. Managed Pages separately support draft, post and scheduling writes gated by consent, Page access and operation approval.
facebook-cli
├── pages # Read references/pages.md first
│ ├── list [--limit N] [--after <cursor>] # Managed Pages
│ ├── access --page-id <id> # Page access diagnostic
│ ├── account-insights --page-id <id> # Page metrics
│ ├── drafts
│ │ ├── create --page-id <id> ... # WRITE — prompts for approval
│ │ ├── show --page-id <id> --draft-id <id> # Read a native draft
│ │ ├── publish --page-id <id> ... # WRITE — same draft; prompts for approval
│ │ ├── edit --page-id <id> ... # WRITE — prompts for approval
│ │ └── delete --page-id <id> ... # WRITE — prompts for approval
│ └── posts
│ ├── list --page-id <id> # Recent posts with metrics
│ ├── get --page-id <id> --post-ids <ids> # Selected post metrics
│ ├── create --page-id <id> ... # WRITE — publish or schedule; prompts for approval
│ ├── reschedule --page-id <id> ... # WRITE — prompts for approval
│ └── cancel-schedule --page-id <id> ... # WRITE — prompts for approval
├── post
│ ├── read (--post-id <id> | --url <url>) # Read by numeric ID/PFBID or canonical post/photo/video URL
│ ├── comments
│ │ └── read --post-id <id> [--limit N] [--after <cursor>] # Read comments (paginated: data[] + paging.cursors.after; post link at summary.post_url)
│ └── reactions
│ └── read --post-id <id> # Reaction summary only: summary.total_count + summary.reaction_counts (no individual reactor list)
├── link-sharing
│ └── decode-url --url <url> # Resolve a share link; returns original_url
├── me # Your Facebook name + profile ID
│ └── friends [--name] [--city] [--hometown] [--work] [--education] [--filter-mode AND|OR] [--json-query <json>] [--birthday-within-days [N]] [--limit N] [--after <cursor>] # Paginated: data[] + paging.cursors.after
├── marketplace
│ ├── search --query "..." [--out <file>] [options] # Search listings; pass the --out file to shopping.resolve_results, do not read it
│ ├── my-listings [--status active|pending|sold|draft] [--limit N] [--after <cursor>] # Your own listings (active + inactive), paginated
│ ├── listing details (--listing-id <id> | --url '<item link>') [--out <file>] # Item details
│ ├── listing create --title "..." --price N --photo /path [options] # Create; goes LIVE only with photos+condition+category+location (location = --latitude & --longitude; no --location flag), else DRAFT
│ ├── listing edit --listing-id <id> --photo /path [options] # Edit a listing (photos uploaded automatically)
│ ├── listing delete --listing-id <id> # WRITE — permanently delete your listing; prompts for approval
│ ├── listing publish --listing-id <id> # WRITE — take a DRAFT live; needs photos+condition+category+location (location = --latitude & --longitude); prompts for approval
│ └── seller-info --listing-id <id> # Seller ratings/info
├── groups
│ ├── details --group-id <id-or-vanity> # Fetch name, About, visibility, history, tags, ordered rules, and member count
│ ├── search [--keywords "..."] [--role connected|admin|admod|any] [--after <cursor>] # Paginated membership listing; keyword + any searches connected/private and public groups
│ └── posts --group-id <id> # Browse posts in a group; --query filters by text
├── events
│ ├── search [--scope connected|discover] [--keywords "..."] [--location "..."] [--latitude N --longitude=-N] [--radius-in-miles N] [--category ...] [--start-date ...] [--end-date ...] [--limit N] [--after <cursor>] # Paginated: data[] + paging.cursors.after; coordinates beat --location, which is only city-accurate
│ └── details --event-id <id> # Read one event; ID from search `id` or `permalink_url`
├── saved
│ ├── list [--type post|video|link|product|reel|event|page] [--collection-id <id>] [--limit N] [--after <opaque>]
│ ├── add --savable-id <FBID> [--type ...] [--collection-id <id>] # WRITE — auto-allowed
│ ├── remove --savable-id <FBID> [--type ...] [--collection-id <id>] # WRITE — auto-allowed
│ └── collections
│ ├── list
│ └── create --name "..." # WRITE — auto-allowed
├── story
│ └── feed [--limit <n>] # Fetch your story feed (story tray)
├── feed
│ ├── newsfeed [--limit <n>] [--after <cursor>] # Ranked algorithmic newsfeed (organic stories)
│ └── friends [--limit <n>] [--after <cursor>] # Ranked friends-only feed
├── timeline
│ └── fetch --profile-id <id> [--limit N] [--after <cursor>] # Fetch timeline posts; use next_cursor to page (check author_id vs owner_id)
└── profile
└── info --profile-id <id> # Look up profile info
What ships with it
15 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- connector_auth_config.yaml 47 B
- manifest.yaml 3.4 KB
- references/comments.md 2.3 KB
- references/events.md 4.8 KB
- references/feed.md 2.6 KB
- references/friends.md 10 KB
- references/groups.md 4.6 KB
- references/marketplace.md 33 KB
- references/pages.md 12 KB
- references/posts.md 4.6 KB
- references/profile.md 2.3 KB
- references/reactions.md 1.9 KB
- references/saved.md 3.7 KB
- references/story.md 942 B
- references/timeline.md 5.0 KB
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.
- 12d ago First seen · 200 lines · 118 tokens per session scan A 619a57e06f37
facebook_cli is a skill published in the GitHub repository asgeirtj/system_prompts_leaks (69,759 stars, last pushed yesterday), licensed CC0-1.0. It adds 118 tokens to every session and 5,364 once invoked, about $0.0005 per session on Opus 5.5. 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 2026-10-05.
Other skills, from other repositories
comps-analysis
Build comparable-company valuation workbooks in Excel.
anatomy
Explain an idea as an interactive isometric machine figure.
pretext
Build creative browser demos with DOM-free text layout.
weights-and-biases
W&B: log ML experiments, sweeps, model registry, dashboards.
creative-ideation
Generate ideas via named methods from creative practice.
nemo-curator
Curate LLM training data: dedupe, filter, PII redaction.