Files
aish/safety.lua
T
marfrit bd59ce7243 safety: is_destructive static pattern matcher (Phase 3 commit #1)
Phase 3 commit #1 per docs/PHASE3.md §12. Static-pattern destructive-op
heuristic; no LLM second-opinion yet (lands in commit #2).

Implementation:
  - 34 patterns in DESTRUCTIVE_PATTERNS table, grouped:
      9 shell-wrapper patterns (R-B1 — bash -c / sh -c / zsh -c / eval /
        python -c / perl -e / pipe-to-sh both forms / pipe-to-bash both
        forms / xargs ... rm). HALT on the wrapper itself; user reads
        the inner before proceeding.
     10 filesystem destructive (rm -rf, find -delete, dd to device, mkfs,
        shred, wipefs, truncate -s 0, ...).
      5 version-control destructive (git push --force/-f, git reset
        --hard, git clean -fd, git branch -D).
      5 database/process (DROP TABLE/DATABASE, TRUNCATE TABLE,
        kill/pkill -9).
      2 permission (chmod 777, chown on root path).
  - ci=true flag for case-insensitive SQL patterns; rule patterns must
    be lowercase when ci is set (matcher lowercases input).
  - pkill -9 ordered BEFORE kill -9; kill rule uses %f[%w] frontier so
    "pkill -9 nginx" reports "pkill -9" not "kill -9" substring match.
  - M._patterns exposes the rule table for :safety patterns meta (Phase
    3 commit #5) and for the test corpus.
  - M.norris_step stub stays — lands in commit #4.

Test corpus (test_safety.lua, 87 cases):
  - 49 destructive cases across all categories (incl. all 11 wrapper
    forms, the canonical curl|sh end-of-string bypass, sudo-prefixed
    rm -rf, etc.).
  - 38 safe cases (read-only commands, non-destructive variants
    of risky verbs like "git push" without --force, "find" without
    -delete, "chmod 644", "kill 1234" without -9, etc.).
  - Documented one accepted false positive: echo "rm -rf /" matches
    the rm pattern by substring — Norris user can proceed after
    reading; tradeoff between false positives and false negatives,
    biased toward false positives per §5.
  - Run from repo root: `luajit test_safety.lua`. Exit 0 on pass.
  - Verified all 87 pass at commit time.

R-C4 / readline rebind, broker opts.max_tokens, LLM second-opinion,
norris_step planner, repl driver, and the wider Norris UX land in
subsequent commits per §12.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-12 22:47:10 +00:00

138 lines
6.7 KiB
Lua

-- safety.lua — workflow safeguards for tool execution.
-- Phase 2: M.confirm_tool_call (per-call confirm gate + auto-approve policy).
-- Phase 3: M.is_destructive (static pattern + LLM second-opinion gate for
-- Norris autonomous mode) and M.norris_step (single-iteration
-- planning loop). See docs/PHASE2.md §6 and docs/PHASE3.md §4 / §5.
local rl = require("ffi.readline")
local json = require("dkjson")
local M = {}
-- Render the call as `name({"path":"/tmp"})` for the confirm prompt.
-- Truncate to keep one-line prompts.
local function pretty_call(name, args)
local body = ""
if args and next(args) then
local ok, encoded = pcall(json.encode, args)
if ok then
body = (#encoded <= 80) and encoded or (encoded:sub(1, 77) .. "...")
else
body = "..."
end
end
return name .. "(" .. body .. ")"
end
-- Ask the user whether tool `name` may be called with `args`, consulting
-- `cfg.mcp.auto_approve` first. Policy keys:
-- "<alias>__<tool>" → exact-match auto-approve
-- "<alias>__*" → whole-server auto-approve
-- Anything else falls back to a [y/N] prompt; empty / non-"y" answer rejects.
-- The separator switched from "." to "__" 2026-05-12 because Anthropic via
-- Bedrock rejects dots in tool names (regex ^[a-zA-Z0-9_-]{1,128}$).
function M.confirm_tool_call(name, args, cfg)
local policy = (cfg and cfg.mcp and cfg.mcp.auto_approve) or {}
if policy[name] then return true end
local alias = name:match("^(.-)__")
if alias and alias ~= "" and policy[alias .. "__*"] then return true end
local prompt = ("call '%s'? [y/N] "):format(pretty_call(name, args))
local ans = rl.readline(prompt) or ""
return ans:lower():sub(1, 1) == "y"
end
-- ---------------------------------------------------------------- is_destructive
-- Phase 3 commit #1: static-pattern matcher only (no LLM second-opinion yet —
-- that lands in commit #2). Patterns are Lua patterns (NOT regex). When
-- `ci = true` is set on a rule, the input is lowercased before matching so
-- the rule matches case-insensitively (`DROP TABLE`, `drop table`, etc.).
-- See docs/PHASE3.md §5 for the rationale and the wrapper-bypass class
-- (R-B1) the first nine entries below are guarding against.
local DESTRUCTIVE_PATTERNS = {
-- ── Shell wrappers (R-B1) — flag the wrapper itself; can't inspect
-- the inner content safely without parsing the inner shell.
-- Norris HALTs on these unconditionally; the user reads the inner
-- before proceeding.
{ pat = "^%s*bash%s+%-l?c%s", reason = "bash -c (wrapped shell)" },
{ pat = "^%s*sh%s+%-l?c%s", reason = "sh -c (wrapped shell)" },
{ pat = "^%s*zsh%s+%-l?c%s", reason = "zsh -c (wrapped shell)" },
{ pat = "^%s*eval%s", reason = "eval (dynamic shell)" },
{ pat = "^%s*python3?%s+%-c%s", reason = "python -c (inline script)" },
{ pat = "^%s*perl%s+%-e%s", reason = "perl -e (inline script)" },
{ pat = "|%s*sh%s", reason = "pipe-to-sh" },
{ pat = "|%s*sh%s*$", reason = "pipe-to-sh (eol)" },
{ pat = "|%s*bash%s", reason = "pipe-to-bash" },
{ pat = "|%s*bash%s*$", reason = "pipe-to-bash (eol)" },
{ pat = "xargs%s+.-rm", reason = "xargs ... rm" },
-- ── Filesystem destructive
{ pat = "rm%s+.-%-rf?", reason = "rm -rf" },
{ pat = "rm%s+.-%-fr?", reason = "rm -fr" },
{ pat = "find%s+.-%-delete", reason = "find -delete" },
{ pat = "find%s+.-%-exec%s+rm", reason = "find -exec rm" },
{ pat = ">%s*/dev/sd[a-z]", reason = "write to raw disk" },
{ pat = "dd%s+.-of=/dev/", reason = "dd to device" },
{ pat = "mkfs%.", reason = "mkfs (format)" },
{ pat = "shred%s", reason = "shred" },
{ pat = "wipefs%s", reason = "wipefs" },
{ pat = "truncate%s+.-%-s%s*0", reason = "truncate to zero" },
-- ── Version control destructive
{ pat = "git%s+push%s+.-%-%-force", reason = "git push --force" },
{ pat = "git%s+push%s+.-%-f%f[%s]", reason = "git push -f" },
{ pat = "git%s+reset%s+.-%-%-hard", reason = "git reset --hard" },
{ pat = "git%s+clean%s+.-%-fd?", reason = "git clean -fd" },
{ pat = "git%s+branch%s+.-%-D", reason = "git branch -D" },
-- ── Database / process
-- ci=true rules use lowercase patterns; the matcher lowercases the
-- input before testing. Don't use uppercase patterns with ci=true.
{ pat = "drop%s+table", reason = "DROP TABLE", ci = true },
{ pat = "drop%s+database", reason = "DROP DATABASE", ci = true },
{ pat = "truncate%s+table", reason = "TRUNCATE TABLE", ci = true },
-- pkill BEFORE kill so the more specific match wins (Lua tables are
-- order-preserving; first hit reports the reason).
{ pat = "pkill%s+%-9", reason = "pkill -9" },
-- kill -9 needs a word boundary so "pkill -9" doesn't match this rule's
-- "kill" substring. %f[%w] is Lua's frontier pattern — matches a
-- transition from non-word to word characters.
{ pat = "%f[%w]kill%s+%-9", reason = "kill -9" },
-- ── Network/permission
{ pat = "chmod%s+.-777", reason = "chmod 777" },
{ pat = "chown%s+.-%s+/%s*$", reason = "chown on root path" },
}
-- Match each rule against `cmd`. Returns (true, reason) on first hit;
-- (false, nil) if no rule matches. Used by the Norris loop to gate
-- shell commands; ALSO called on tool-call args by Norris's tool path
-- (the JSON-serialized arguments are passed in as cmd).
function M.is_destructive(cmd)
if type(cmd) ~= "string" or cmd == "" then return false, nil end
local lower = nil -- lazily computed for ci-rules
for _, rule in ipairs(DESTRUCTIVE_PATTERNS) do
local target = cmd
if rule.ci then
lower = lower or cmd:lower()
target = lower
end
if target:match(rule.pat) then
return true, rule.reason
end
end
return false, nil
end
-- Expose the pattern table for `:safety patterns` meta and for testing.
M._patterns = DESTRUCTIVE_PATTERNS
-- ---------------------------------------------------------------- norris_step
-- Phase 3 commit #4 lands the planner. Stub stays for now.
function M.norris_step(plan, broker, executor)
error("safety.norris_step: not implemented yet (lands in Phase 3 commit #4)")
end
return M