badger/docs/ at main · jenchan.biz/badger · Tangled

jenchan.biz / badger

0

Prompt-injection defenses for LLM agents — fetch interception, output sanitization, session inspection.

0

badger / docs /

at main 1 folder 2 files

Jen Chan refactor(inspector): consolidate utilities into utils module 3w ago

docs: record canary detection plan + worktree include config - CONTEXT.md: define Canary + canary_leak, replacing the prompt_revealed overclaim (badger can't plant into the host's real system prompt) - docs/adr/0001: ADR for canary-detects-a-planted-marker, silent+opaque, persisted-on-disk bounded risk, host asymmetry - .worktreeinclude + .gitignore: copy local config/scratch (.env, docs/issues, docs/plans, .claude/settings.local, .badger) into isolated agent worktrees; ignore .claude/worktrees - naming.md: session-inspector → inspector; opencode.json: add fetch / approve_llm_analysis commands Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

3 weeks ago

dataflow-sequence.md

badger: prompt-injection defenses for LLM agents pnpm monorepo, published as @usrrname/badger-*: - badger-fetch: MCP server + Claude/OpenCode hooks that intercept agent web fetches, guard requests (HTTPS/SSRF/rate-limit), and enforce a two-signal pass/neutralize/block policy on responses - badger-response: zero-dependency library detecting prompt-injection compromise, exfiltration, and obfuscation in tool output - badger-inspector: terminal UI for Claude Code / OpenCode session transcripts, flagging compromise inline Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

1 month ago

webfetch-flow.md

badger: prompt-injection defenses for LLM agents pnpm monorepo, published as @usrrname/badger-*: - badger-fetch: MCP server + Claude/OpenCode hooks that intercept agent web fetches, guard requests (HTTPS/SSRF/rate-limit), and enforce a two-signal pass/neutralize/block policy on responses - badger-response: zero-dependency library detecting prompt-injection compromise, exfiltration, and obfuscation in tool output - badger-inspector: terminal UI for Claude Code / OpenCode session transcripts, flagging compromise inline Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

1 month ago