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/endogenai/dogma/source-cachingnpx skills add EndogenAI/dogma --skill source-cachinggit clone --depth 1 https://github.com/EndogenAI/dogmaWhat 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 | $0.00121 | $0.01174 |
| Opus 5 | $0.00060 | $0.00587 |
| Sonnet 5 | $0.00024 | $0.00235 |
| Haiku 4.5 | $0.00012 | $0.00117 |
Grade A, and why
source-caching 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 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.
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 — 125 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Source Caching
This skill enacts the Local Compute-First axiom from MANIFESTO.md: populate the source cache locally first, then research sessions read from disk rather than re-fetching through the context window. The fetch-before-act posture is governed by AGENTS.md § Programmatic-First Principle and § Security Guardrails. When this skill and those documents conflict, the primary documents take precedence.
1. Pre-Warm at Session Start
At the beginning of any research session, warm the entire source cache before delegating to any scout:
# Dry run first — see what will be fetched without fetching
uv run python scripts/fetch_all_sources.py --dry-run
# Fetch all uncached sources (idempotent — skips already-cached URLs)
uv run python scripts/fetch_all_sources.py
This batch-fetches all URLs from OPEN_RESEARCH.md and existing research doc frontmatter. Scouts then read cached Markdown files with read_file rather than consuming tokens on live web fetches.
2. Check Before Fetching Any Individual URL
Before fetching a specific URL, always check whether it is already cached:
uv run python scripts/fetch_source.py <url> --check
# Exit 0 = cached; Exit 2 = not cached
Never re-fetch a cached source without --force — it wastes tokens and rate-limit budget.
3. Fetch and Cache a Single URL
uv run python scripts/fetch_source.py <url>
# Saves distilled Markdown to .cache/sources/<slug>.md
# Prints the slug and cache path on success
To force-refresh an already-cached page:
uv run python scripts/fetch_source.py <url> --force
4. Get the Cached File Path
After caching, retrieve the path for use with read_file:
uv run python scripts/fetch_source.py <url> --path
# Prints: .cache/sources/<slug>.md
Use this path with the read_file tool to read the distilled content without re-fetching.
5. List All Cached Sources
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 · 125 lines · 121 tokens per session scan A 03abc1a9049f
source-caching is a skill published in the GitHub repository EndogenAI/dogma (2 stars, last pushed 8d ago), licensed Apache-2.0. It adds 121 tokens to every session and 1,174 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 2026-08-31.
Other skills, from other repositories
database-design
Design schemas and migrations when persistent data structures change.
dispatching-parallel-agents
Evaluate when to dispatch parallel agents; use a 4-step pattern to split, coordinate, and integrate results.
verification-before-completion
Enforce "no evidence = no completion"; run Gate Function verification before declaring done.
test-driven-development
Drive changes with Red→Green→Refactor; ensure behavior is verifiable and regression-safe.
doc-lookup
Check official docs when framework APIs or configuration may be version-sensitive.
air-blackbox-sales-agent
AIR Blackbox's autonomous sales prospecting agent. Finds Python AI projects on GitHub that need EU AI Act compliance, identifies the right person to contact (CEO, CTO, lead maintainer), runs a free compliance scan, and drafts personalized outreach emails that convert to engagement. The sales flow: free scan as the…