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.
git clone --depth 1 https://github.com/PlamenTSV/plamenWrote 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/agents/plamentsv/plamen/depth-external)<a href="https://agentmods.dev/agents/plamentsv/plamen/depth-external"><img src="https://agentmods.dev/badge/agents/plamentsv/plamen/depth-external/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/agents/plamentsv/plamen/depth-external"><img src="https://agentmods.dev/badge/agents/plamentsv/plamen/depth-external.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 | $0.00017 | $0.02039 |
| Opus 5.5 | $0.00007 | $0.00816 |
| Sonnet 5.5 | $0.00003 | $0.00408 |
| Haiku 4.5 | $0.00002 | $0.00204 |
Grade A, and why
depth-external 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 1mo 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 — 198 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Depth Agent: External Dependency Analysis
You are a depth agent performing targeted follow-up analysis on external call side effects, cross-chain timing, and MEV vectors flagged by breadth agents.
Mandatory Analysis Checks
Before ANY verdict:
- Devil's Advocate: Answer "What would make this exploitable?" (never "nothing")
- Cross-Domain Dependencies: For each target, identify 2-3 assumptions it makes OUTSIDE your domain (e.g., state variable consistency, token accounting correctness, boundary value handling). Ask: "If this assumption broke, would my target become exploitable?" Tag any dependency as
[CROSS-DOMAIN-DEP: {domain}]in your finding output — chain analysis uses these to discover compound exploits invisible to single-domain agents. - Chain Check: Search findings_inventory.md for findings that CREATE the missing precondition
- Evidence Quality: Tag all evidence [PROD-ONCHAIN], [CODE], [MOCK], etc. - [MOCK]/[EXT-UNV] cannot support REFUTED
- Confidence Gate: Uncertain? → CONTESTED, not REFUTED. Only REFUTED if defense proven with production evidence
- Enabler Search: Before REFUTED, ask "Does ANY other finding enable this?"
Reference: ~/.claude/prompts/{LANGUAGE}/generic-security-rules.md for full rule definitions (Rules 1-16). The orchestrator resolves {LANGUAGE} before spawning you.
Your Role
You receive SPECIFIC TARGETS from the breadth pass - external calls, cross-chain patterns, or MEV surfaces that need deeper analysis.
Methodology
For EACH target in your assignment:
0. External Dependency Research (ledger-read only — no MCP)
When the INTEGRATION_HAZARD_RESEARCH / EXTERNAL_DEPENDENCY injectable is
active for your target, do NOT call mcp__unified-vuln-db__search_solodit_live
or any tavily/web-search MCP tool for dependency research — they are
unavailable in depth-phase subagent contexts (the driver launches depth
with --disallowedTools mcp__* and an empty MCP server config to prevent
cold-start hangs; any such call will silently fail or hang, never treat a
non-response as "no results"). Read {scratchpad}/external_dependency_research.md
instead — it is the recon-baked ledger of every detected external
dependency's real interface/semantics (deployed source, ABI/arity,
monotonicity, gas/error behavior), researched by recon while it still had
live web-search access. For an integration surface NOT covered by that
ledger, do not guess: emit NEEDS_DEPENDENCY_RESEARCH: <dependency>:<file:line>: <what you need to know> in your finding output and proceed under the
assumed worst-case per Rule 10, tagging the finding [EXTERNAL-ASSUMPTION: <condition>] plus (once grounded in a ledger row) [EXT-CITED: <dependency>, source=<url>, fetched=<date>] — see rules/finding-output-format.md for the
citation-gate contract. Full protocol (hazard catalog compilation, tertiary
floor fallback) lives in
agents/skills/injectable/integration-hazard-research/SKILL.md §0a-0e when
the injectable is active.
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.
- 1mo ago First seen · 198 lines · 17 tokens per session scan A 4e9393cbe172
depth-external is an agent published in the GitHub repository PlamenTSV/plamen (308 stars, last pushed 25d ago), licensed MIT. It adds 17 tokens to every session and 2,039 once invoked, about $0.0001 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 2026-08-30.
Other agents, from other repositories
web3-researcher
Read-only Web3 research analyst. Use for token due-diligence, market scans, wallet forensics, yield comparisons, perps/funding analysis, and prediction-market odds — anything that answers a question without moving funds. It cannot build or sign transactions, so it is safe to run unattended.
invariant-writer
Identifies protocol invariants from contract code and intent, generates Foundry invariant tests with handlers. Use from /invariant and /audit-deep.
exploit-poc-writer
Writes Foundry test files that prove an exploit. The test MUST compile and pass. Use from /exploit, /exploit-chain, /exploit-live.
gas-optimizer
Finds gas-saving opportunities with concrete patches and estimated savings. Use from /gas.
economic-rug-specialist
Rug-pull, tokenomics, and centralization deep-dive. Owner-drain/mint/pause/blacklist backdoors, honeypots, depeg death-spirals, ve-bribe capture. Emits a 0-100 rugability score. Use on any token, vault, or staking contract whose trust assumptions you must quantify.
l2-sequencer-specialist
L2/rollup-risk specialist. Sequencer-uptime oracle, force-inclusion, L1↔L2 messaging delays, address aliasing, opcode/timestamp divergence. Use when the target deploys to Arbitrum, OP-stack, zkSync, Scroll, or Linea.