agent_unbounded_resources

Bound the agent loop with an explicit iteration / token / time budget low effort

# Before (VULNERABLE): unbounded paid completions
while True:
    resp = llm.invoke(state)
    state = step(resp)

# After (SAFE): a hard turn cap + a token budget
for _ in range(MAX_TURNS):
    resp = llm.invoke(state, max_tokens=512)
    state = step(resp)
    if state.done:
        break

Treat agent budgets like rate limits — never unlimited low effort

Suppress a confirmed non-issue with a committed .pullguardignore entry (pullguard ignore locally, or comment /pullguard ignore <fingerprint> <reason> on the PR — the fingerprint is printed in the PR comment’s Triage section). Entries support expiresAt for time-boxed snoozes.

Security findings at major or critical severity — and any critical finding — always surface: .pullguardignore cannot hide them. The reviewed paths that keep them visible are acknowledged (reviewed, stays in reports) and, for a confirmed false positive, a reasoned false_positive entry — visible and audited, excluded only from the merge block.