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 skills/stevepridemore/graph-memory/graph-capturenpx skills add stevepridemore/graph-memory --skill graph-capturegit clone --depth 1 https://github.com/stevepridemore/graph-memoryWhat 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 | $0.00063 | $0.02722 |
| Opus 5 | $0.00032 | $0.01361 |
| Sonnet 5 | $0.00013 | $0.00544 |
| Haiku 4.5 | $0.00006 | $0.00272 |
Grade C, and why
graph-capture scanned grade C 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 2d 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.
Harvests environment variableshighData exfiltration
Enumerating or grepping the environment for keys collects credentials unrelated to what the mod says it does.
5. **Never extract secrets.** API keys, passwords, tokens, connection strings, private keys, signed URLs — none of these go into entity properties or edge evidence. Note that a credential exists ("Project X uses an API k How it starts
The opening of the file, as written. The whole thing — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
The user wants to capture the current conversation to the memory graph. Many conversations contribute knowledge that the dream process can't see (claude.ai web and Desktop chats live server-side, not in the local transcript store the dream walks). This command is the manual catch-up: review what we just talked about, find what hasn't already been written to the graph, and write it.
Arguments: $ARGUMENTS
Parse:
--dry-run— describe what you would write without calling write tools. Useful to review before committing.--topic <focus>— focus capture on entities/relationships related to a specific topic mentioned in the conversation. Skip unrelated material. Default: capture everything substantive.--since-message <uuid>— only consider messages from that point onward. Default: whole conversation.
Steps
1. Inventory candidate entities and relationships
Walk the conversation (or the slice indicated by --since-message / --topic) and list every distinct candidate that meets the "worth writing" bar:
- People named with role, organization, or relationship context — not just casual references
- Projects worked on, evaluated, or referenced with meaningful context
- Technologies / Concepts the user used, evaluated, decided about, or expressed a preference toward
- Preferences explicitly stated ("I prefer X", "always use Y") or strongly implied through repeated choice
- Decisions made with reasoning ("we decided X because Y") — both explicit and clear inferred decisions
- Facts about infrastructure, processes, configuration, or the user's environment
- Events — meetings, deployments, incidents, milestones with dates or outcomes
- Objects — specific repos, servers, databases, tools, containers
- Reasoning traces — how a problem was solved, especially if there were dead ends or alternatives considered. Capture the trace, not just the outcome.
Skip:
- Conversation mechanics ("let me read that file", "running the test now")
- Trivial mentions without significance
- API keys, passwords, tokens, connection strings, or any secret value (note existence only, never the value)
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.
- 2d ago First seen · 143 lines · 0 tokens per session scan C c52f6d8355cc
graph-capture is a skill published in the GitHub repository stevepridemore/graph-memory (2 stars, last pushed 3mo ago), licensed MIT. It adds 63 tokens to every session and 2,722 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 1 finding (harvests environment variables). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ori-memory
Persistent agent memory with learning retrieval. Knowledge graph on markdown files — capture insights, decisions, research, and learnings during work, then retrieve them weeks or months later. Use when knowledge is too valuable to lose but too much to inject into every prompt.
svoboda-profiler
Use when asked to profile someone, unpack a person, build a psychological portrait, run a Svoboda session, or map life state. Triggers on распаковка, профилирование, психопортрет, точка А, колесо баланса, profile someone, unpack, svoboda session.
vault-scaffolder
Use when a completed svoboda profile.yaml exists and a new Personal OS needs to be personalized in the current workspace.
install-waggle
Install the official Waggle plugin locally for Codex from the latest stable Abhigyan-Shekhar/Waggle-mcp GitHub Release. Use for explicit requests such as "Install Waggle", "Set up Waggle", "Install Waggle for Codex", or "Enable Waggle". Do not use for Waggle operation, repair, updates, configuration, or memory…
lint
Use weekly or after major ingest sessions to check wiki health.
start
One-word self-unfolding onboarding. Sets up a fresh, unpersonalized vault end-to-end — installs the memory engine, runs the profile interview, scaffolds the vault, and teaches the user as it goes. The single front door.