Commit Graph

7 Commits

Author SHA1 Message Date
gsxdsm
189f237a07 fix(security): make the approval and permission gates actually enforce (#2457)
## Why

An AI agent autonomously deleted a live task (FN-8609) via
`fn_task_delete`. Investigating that led to two audits, which found the
gate machinery is well-built and its *enforcement* well-tested — but the
*decision boundary* was unguarded and the tools that matter never
reached the gate at all.

The headline defect: all ~76 tools in `packages/cli/src/extension.ts`
are delivered through `hostExtensionPaths` → `DefaultResourceLoader`,
which never passes through `wrapToolsWithActionGate`. `fn_task_delete`
*was* classified as gated in `gating-classifications.ts` — that
classification was dead code. Roughly 15 destructive tools were ungated
even for an operator who had locked their policy all the way down.

Second: `POST /api/approvals/:id/decision` took the approver identity
from the request body, validated for shape only, with no check that the
decider was human or differed from the requester. Combined with a bash
tool whose command string was never inspected (only `params.cwd`), an
agent could read the daemon token from `~/.fusion/settings.json`, find
its own pending request, and approve it.

## Scope

This does **not** tighten defaults. The `unrestricted` preset is
byte-for-byte unchanged — `git diff` on `agent-permission-policy.ts` is
empty — and regression tests assert that an out-of-the-box install
behaves exactly as before. The bug was never "the default is too
permissive"; it was "strict policy doesn't enforce." This makes turning
security up actually work.

The one deliberate exception: the containment that stops an agent
escalating its *own* privileges (reading the daemon token / credentials,
calling the approvals API to self-approve) applies at every preset
including `unrestricted`. That is a privilege-escalation boundary rather
than a permission preference — if it only engaged under strict policy it
would not have prevented the incident that prompted this.

## What changed

8 bisectable commits:

- **Approval lifecycle** — self-approval blocked via server-derived
deciders; same-verdict replay 409s; decide re-reads and re-validates
inside the transaction; expiry TTLs; `markCompleted` ownership check;
session identity registry in core.
- **Engine gates enforce for real** — unclassified tools resolve to a
policy-governed category instead of hardcoded `allow`; missing-policy
fail-open closed; bash containment floor + exact-command approval
binding.
- **Dashboard decision routes** — stop trusting client-supplied actors
(decision, bypass-review, worktrunk → 403 on forged actors).
- **`fn serve` authenticated by default** — auto-mints a token following
the existing `fn dashboard` precedent; `--no-auth` opts out.
- **Sibling entry points closed** — user-sourced hard-cancel moves, ACP
execute-once approvals, plugin task-store gating.
- **pi-extension principal resolution** — the extension resolves the
acting principal and can withhold or policy-gate the previously ungated
destructive tools.
- **Root-cause bonus fix** — `findLatestByDedupeKey` was broken in
PostgreSQL backend mode (already-parsed jsonb fed through a string-only
parser), so approved-grant redemption **never matched in production**,
minting duplicate requests. This explains the live DB state of 17
approved / 0 completed. *(Also cherry-picked to `main` as `a9b30013bb`,
since it is an active production defect on its own.)*
- **Review follow-ups** (`627f1b1fa8`) — operator-configured
provisioning privilege and a configurable grant TTL; see below.

## Review follow-ups

**Provisioning privilege is operator-configured, not role-derived.**
`isCallerPrivileged` had gone from `caller.reportsTo == null` (every
top-level agent privileged — permanent escalation by creating a
manager-less agent) to `caller.role === "ceo"`, which swapped an
implicit rule for a magic string: any agent config can claim that role,
while an operator who genuinely wants a privileged agent had no
supported way to say so. Privilege now derives solely from
`agentProvisioning.trustedAgentIds` / `trustedRoles` and fails closed
when settings are unresolvable.

It is also no longer forwarded to `resolveAgentProvisioningPolicy` as
`isPrivileged`, because that flag short-circuits ahead of
`alwaysApproveDelete` — a trusted caller was bypassing delete approval
entirely. The policy applies the same trusted rules itself, in the right
order. The function now governs only the org-chart escape hatch (acting
outside your own direct reports).

**Grant TTL defaults to 1 hour and is configurable.** Approval →
redemption is not instantaneous: an operator approving from their phone,
an engine restart, a queued lane, or a task waiting on a worktree all
routinely exceeded 15 minutes, after which the grant expired and the
agent silently re-requested. One hour remains far short of the
"redeemable forever" hazard the TTL exists to bound. Override via
`FUSION_APPROVAL_GRANT_TTL_MS` or `configureApprovalRequestTtls()`;
invalid overrides are ignored rather than widening the window to
infinity or collapsing it to zero.

## Behavior changes requiring operator review before rollout

1. `fn serve` requires a bearer token by default (`--no-auth` opts out);
unauthenticated clients get 401.
2. Agents can no longer run withheld destructive tools
(`fn_task_delete`, `fn_task_bypass_review`,
mission/milestone/slice/feature/workflow deletes, `experiment_finalize`,
`skills_install`). Operators keep them via CLI/dashboard. **This is the
incident fix.**
3. Agents get provisioning privilege only when the operator lists them
in `agentProvisioning.trustedAgentIds` / `trustedRoles`; the
provisioning gate is now live in production. Previously-implicit
privilege (top-level position, or a `ceo` role) no longer grants
anything on its own.
4. Decision replay 409s (was 200); pending approvals expire after 24h,
approved grants after 1h (configurable); bash approvals bind per exact
command.
5. Forged/body actors on decision, bypass-review, worktrunk routes →
403; `archive-all-done` requires `{confirm:true}` (external scripts
affected).
6. `fn_secret_get` approvals grant exactly one reveal (previously
granted nothing and looped forever); ACP approvals are execute-once
(previously infinite reuse).
7. Bash containment denies token/credential/approvals-API commands in
all agent sessions at every preset.

## Verification

Independently re-run against the branch, not just self-reported:

- 5 typechecks (core, engine, cli, dashboard `tsconfig.json` +
`tsconfig.app.json`) — clean
- `pnpm lint` — clean
- `pnpm test:gate` — 379 passed
- `pnpm build --force` — green (a plain `pnpm build` skips packages as
unchanged and does **not** compile the branch)
- `pnpm check:changesets` — clean
- ~650 file-scoped tests including new negative-path suites for the
decision boundary, which previously had **zero** test coverage

`packages/engine/src/__tests__/plugin-runner.test.ts` fails 56/80 —
**verified pre-existing**, reproducing identically at base commit
`93a403af67` on `main`. Not in the merge gate.

### A mutation check that failed to fail

Worth recording, because it nearly shipped an untested security fix. The
first mutation check on the provisioning change reintroduced the `ceo`
hardcode and **all 17 tests still passed** — the tests asserted through
the policy path, which can no longer observe `isCallerPrivileged` at
all, precisely because `isPrivileged` is no longer forwarded there.
Org-chart cases that do exercise the function were added; the hardcode
now fails exactly 1 of 19, and restoring is green. A green mutation run
is only meaningful if the test can actually see the code under test.

## Known limitations (stated, not papered over)

- The bash containment floor is string-matching: a cost-raiser, not a
sandbox. Quoting, encoding, `$HOME`, symlinks, or an interpreter
one-liner can evade it. The durable protection is the decision route
refusing agent-originated deciders — the filter is the belt, not the
braces.
- Approval expiry is lazy (evaluated at decide/complete/redeem), not
swept, so an expired pending row stays visible in lists until touched.
- The extension's require-approval path returns a pending message but
cannot suspend a pi session mid-turn; engine-side pause hooks cover
engine lanes only.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Security**
* Hardened approval and permission gating with server-side decider
attribution, self-approval blocking, ownership checks, replay/race
protection, and status/TTL enforcement.
* Added fail-closed behavior for sensitive/unclassified tools and
sandbox provisioning approvals.
* Blocked credential/approval access via bash containment; plugin
destructive task operations now require explicit permission.
* **New Features**
* `fn serve` now defaults to bearer-token auth, with `--no-auth` as the
explicit opt-out.
* **Bug Fixes**
* Improved task move-source attribution (`moveSource: "user"`) and
tightened dashboard archive/bypass confirmation and operator attribution
behavior.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-07-30 21:50:37 -07:00
gsxdsm
d3e1a355dc fix(acp): plan output charged to per-turn budget; HITL requires readable decision
Two valid P1s from PR review threads:
- event-bridge: plan output bypassed the per-turn output cap — entry size was
  bounded but entry COUNT wasn't (1000 entries ~ 64MB through onThinking).
  Plans are now suppressed once the cap flags, capped at MAX_PLAN_ENTRIES=100
  with a truncation marker, bounded, and charged to the budget. +2 tests.
- control-handler: with pauseForApproval but no findApprovalByDedupeKey, a
  human approval was silently discarded (unreadable status -> deny). HITL now
  requires BOTH closures upfront and default-denies before creating a request,
  so no approval is wasted and no pending record orphaned.

184 tests green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 12:56:58 -07:00
gsxdsm
b9d824b41f fix(acp): address bot review feedback on PR #1354
- Check pauseForApproval BEFORE createApprovalRequest so a gate with
  create-but-no-pause default-denies without orphaning a pending approval
  record (greptile P1).
- prompt-builder: whitespace-only prompt yields no text block (code/comment
  mismatch) + regression test.
- onLoad logs arg count, not raw args (args can carry inline tokens).
- Document that engine-driven session resume (loadAcpSession) is deferred v1.
- Strengthen tests: eviction path observed end-to-end, id-normalization
  asserted via differing raw forms, loadSession receives the normalized id.

Skipped with reasons (recorded in review thread reply): exports-to-dist,
README title (package name is correct), Surface Enumeration boilerplate,
heavy-lift streaming-read/path-jail rework, and two suggestions that would
weaken the default-deny floor. 182 tests green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 12:30:28 -07:00
gsxdsm
63d96bd418 refactor(acp): simplify pass — wire exit-hook, drop dead idle timer
Post-implementation /simplify cleanup. Registers the plan-specified
process.on('exit', killAllProcesses) safety hook in index.ts (was missing —
closes an orphan-subprocess gap on hard exit). Removes the unwired idle-timer
(engine StuckTaskDetector + dispose()/registry teardown is authoritative per
KTD4a) and its tests. Fixes a stale dispositionFor doc comment, removes a
redundant identifier re-normalization in the event bridge, and clarifies why
the FusionCategory type keeps git_write/task_agent_mutation. Behavior-
preserving; 177 tests green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 09:31:32 -07:00
gsxdsm
489a287d6f feat(acp): bundle into CLI, on-demand install, S1 safety + evidence (U8)
Wires the ACP runtime plugin into the published CLI (RUNTIME_PLUGIN_IDS in
tsup.config) and the on-demand BUILTIN_PLUGINS catalog (experimental), matching
the untrusted-subprocess security posture. Adds the Risk S1 default-policy
safety: an acpAllowUnrestricted acknowledgement (default false) — without it, a
blanket allow on a sensitive category is escalated to approval rather than
auto-approved under the allow-all default policy, applied in both the permission
floor and fs write gating. Adds docs/acp-contract.md (launch/readiness +
failure taxonomy), a README with the AGENTS.md-required upstream evidence
(SDK repo/docs/release/integrity), a bundle-output test for the staged plugin,
and a @runfusion/fusion minor changeset. Package green at 179 tests.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 09:25:56 -07:00
gsxdsm
d9f1392b86 feat(acp): fs capabilities behind a realpath path-jail (U7)
path-jail.ts is a real symlink-resolving confinement jail (NOT the
project-root-guard string check): realpath validation within realpath(cwd),
parent-realpath + final-component lstat for new files (rejects dangling/symlink
finals), O_NOFOLLOW open + re-validation for TOCTOU, NUL/escape rejection, and
a deny-list for secrets (.env/*.pem/*.key/.npmrc/.netrc/id_*/credentials) and
git internals. fs-capabilities.ts: read honors line/limit + a hard byte
ceiling; write is default-OFF, size-capped, hard-rejects .git/**, and routes
through the file_write_delete gate (reusing the U5 floor) — block/require-
approval gate the write, never free. Handlers registered only when the
capability is enabled, consistent with the advertised fs capability. +39
tests (173 total), incl. real symlink-escape and .git-write rejections.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 09:17:13 -07:00
gsxdsm
7c8a3d4075 feat(acp): per-category permission floor + HITL + cancel drain (U5)
The security floor for session/request_permission. Classifies each tool
call's kind into a Fusion action category and reads the per-category
disposition from the live policy (never a preset shortcut — S1/KTD3a), so a
custom rule blocking command_execution is honored even under the default
unrestricted preset. Selects allow_once only, never allow_always (S2).
Unmappable/missing/other kind and missing gate/policy default-deny;
require-approval routes through the gate's HITL closures (createApprovalRequest
-> pauseForApproval -> re-read status) or default-denies when no approver
exists. requestPermission tracks in-flight requests and drains them cancelled
on teardown (KTD4a). Couples only to a local PermissionGate (no @fusion/engine
import, KTD3). +29 tests (106 total).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-03 09:04:36 -07:00