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/sms021/runawaycontext/runaway-contextnpx skills add sms021/RunawayContext --skill runaway-contextgit clone --depth 1 https://github.com/sms021/RunawayContextWrote 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/sms021/runawaycontext/runaway-context)<a href="https://agentmods.dev/skills/sms021/runawaycontext/runaway-context"><img src="https://agentmods.dev/badge/skills/sms021/runawaycontext/runaway-context.svg" alt="Measured on agentmods" 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 | $0.00111 | $0.01378 |
| Opus 5 | $0.00056 | $0.00689 |
| Sonnet 5 | $0.00022 | $0.00276 |
| Haiku 4.5 | $0.00011 | $0.00138 |
Grade A, and why
runaway-context 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 — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.
RunawayContext Skill
This skill activates the RunawayContext v3 retrieval surface inside Claude Code. When the skill is loaded, the AI has access to:
- Read tools (via the MCP server):
get_brief,search_chunks,search_lessons,list_drafts,list_specialists,brief_preview,audit_verify. - Write tools (via the MCP server):
propose_lesson_draft,approve_draft,reject_draft,regen_brief,brief_rollback,mature_lesson.
When to invoke
-
Entering a new project directory. The AI should call
get_briefwith the project's slug. The brief is auto-generated, capped at 150 lines, and reflects the current curated state. -
The user asks for context. Examples:
- "What do you know about the accounting project?" →
get_brief(slug='accounting'). - "Search for paycom-related lessons." →
search_lessons(query='paycom', project='paycom'). - "What's the data map for Procore?" → query
data_sourcesanddata_source_mappingsforsystem='procore'.
- "What do you know about the accounting project?" →
-
The user corrects you on something durable. When the correction is project-specific and likely to recur, propose a draft:
propose_lesson_draft(title=..., what_happened=..., the_fix=..., prevention_rule=..., project_tags=[...]).- Do not call
log_lessondirectly. The drafts inbox is the human-approval gate (HR-9).
-
The user invokes a slash command:
/runaway:brief <slug>→get_brief(slug)./runaway:search <query>→search_chunks(query)+search_lessons(query)and present merged results./runaway:project <slug>→ switch the active project context (multi-project session stacking)./runaway:drafts→list_drafts()and present for review./runaway:regen <slug>→regen_brief(slug)./runaway:stats→ runrunaway statsand present the output.
Hard rules to honor
This skill enforces the same hard rules as the underlying install:
- HR-1. No network calls. All MCP traffic is stdio. Semantic embeddings are local-onnx unless the install opted in to remote providers.
- HR-2. Every
propose_lesson_draftcall must include at least oneproject_tagsentry from the canonical slug registry. If the user gives you content without a slug, ask for one before drafting. - HR-3. Never call hard-delete via this skill. The MCP server does not expose it.
- HR-9. Never call
mature_lessonwithout explicit user authorization. The engine proposes; the user (or you, under their authority) approves. - HR-13. Do not generate
TODO/FIXMEmarkers when writing files in a RunawayContext install's source tree.
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 · 116 lines · 111 tokens per session scan A 5dba83b16049
runaway-context is a skill published in the GitHub repository sms021/RunawayContext (16 stars, last pushed 3mo ago), licensed MIT. It adds 111 tokens to every session and 1,378 once invoked, about $0.0006 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 2026-08-30.
Other skills, from other repositories
omega-memory
Persistent memory for AI coding agents. Teaches agents how to use OMEGA's MCP tools for storing decisions, querying context, coordinating multi-agent workflows, and resuming tasks across sessions.
omega-memory
Persistent memory for AI coding agents. Semantic search, auto-capture, checkpoint/resume across sessions.
engram
Read and write the project's durable reasoning memory (decisions, principles, cautions, problems, insights) through the Engram MCP tools. Recall relevant memory before non-trivial work; capture only durable, high-value knowledge silently at natural stopping points; keep the graph honest (judge suspects, close answered…
engram-digest
Seed or top up this project's Engram graph from the existing codebase — an explicit, user-invoked digestion of the current working tree into typed memory nodes (decisions, principles, cautions, problems, insights, intents). Use when the user says "digest this project", "ingest the codebase into memory", "seed the…
memseek-feedback
Record success, failure, or a correction against the Memseek context used for the latest request.
memseek-status
Diagnose the Memseek Claude Code connection, project identity, MCP tool package, and queued writes.