security: narrow skill grants, ship ask/deny rules, guard skill frontmatter, GET-only 403 retry (#642)

* security(skills): drop bare Bash and the portal CLI glob from skill grants

A skill's allowed-tools pre-approves the listed tools for the turn that
invokes it, even in an untrusted folder. job-application-assistant granted
bare Bash, Edit and Write; job-scraper granted bun run on any
.agents/skills/*/cli/src/cli.ts. Every Bash command the skills document is
already allowlisted per command in .claude/settings.json, so the bare and
glob grants bought nothing but exposure. The per-portal grants stay in
settings.json, which is where a portal added by /add-portal registers too
(Step 5 now says so).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* security(settings): deny exfil-shaped shell and credential reads, always ask on permission-defining files

Deny rules block in every permission mode including bypassPermissions and
override any allow rule or skill grant; ask rules force a prompt even when
a skill's allowed-tools would pre-approve the call. settings.json now
ships both: deny for shell that uploads or opens a remote shell (curl
data/form/upload flags, wget post, scp, sftp, ssh, nc) and for reads of
.env and ~/.ssh, ~/.aws, ~/.gnupg; ask for the files whose content is the
permission surface (settings, skill frontmatter, agents, CI, this guard)
and for git push. security_guards.py requires each rule to stay present,
the same way it requires the personal-data gitignore rules. Prefix rules
are tripwires, not walls; SECURITY.md will say so.

Also allowlists tools/robots_check.py and the upcoming tools/fetch_page.py
so the 403 retry no longer needs a curl grant.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* security(guards): hold skill and command allowed-tools to a reviewed allowlist

The guard allowlisted settings.json permissions but never read skill
frontmatter, so the narrow settings allowlist sat next to a bare Bash
grant it could not see. Check 4 parses every .claude/skills/*/SKILL.md,
.agents/skills/*/SKILL.md and .claude/commands/*.md frontmatter with a
stdlib parser (scalar, folded, and list forms) and holds each
allowed-tools entry to ALLOWED_SKILL_TOOLS; a portal skill may
additionally grant exactly its own cli/src/cli.ts, matched by path so no
portal name is hardcoded. Bare Bash and mid-pattern wildcards get their
own message. On master's frontmatter this check reports two failures.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* feat(tools): add GET-only fetch_page.py for the browser-header retry

09-web-research.md documented the 403 retry as a curl command with browser
headers plus an inline python3 -c tag-stripper, which meant the workflow
wanted a Bash(curl:*) grant - and curl with a free tail is an upload
channel. fetch_page.py is the same retry as one command with a fixed GET
argv (URL after --, bounded to 45 s / 5 redirects / 10 MB), http(s) only,
robots_check.gate() enforced with no bypass flag, --out refused inside
the repo, and --text for the stripped page. urllib fallback when curl is
absent. The settings.json grant names the script.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs(security): describe the permission surface as a table; route the 403 retry through fetch_page.py

SECURITY.md now lists every shipped grant with its location, lifetime
and guard, and says plainly what the permission layer does not do: no
sandbox on native Windows, prefix rules are tripwires, WebFetch is
unrestricted by domain, subagents run with the parent's tools (next
change), and Edit/Write stay granted because tailoring a CV is editing
files. 09-web-research.md's retry section is the one fetch_page.py
command (framework_version 1.1.3 -> 1.1.4); apply, rank and scrape point
at it instead of curl. SETUP.md and the README tools tree follow.
CHANGELOG gains a Security section under [Unreleased].

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* security(review): ask on command frontmatter, portal skills grant only their CLI, bound the urllib fallback

Findings from a fresh-context review of the branch:

- A command's frontmatter is a grant too (check 4 already parses it), so
  Edit(.claude/commands/*.md) joins the required ask rules; without it a
  steered session could write `allowed-tools: Bash` into apply.md with no
  prompt and the next /apply would be pre-approved for bare Bash.
- SECURITY.md promised a portal skill may grant "its own CLI and nothing
  else"; the guard now enforces exactly that instead of own-CLI plus the
  core allowlist. All six shipped portal skills already comply.
- fetch_page.py's urllib fallback now matches curl's 5-redirect cap and
  applies the 10 MB cap after decompression, so the docstring's bounds
  hold on both paths.
- Frontmatter parser: a BOM no longer reads as "missing frontmatter",
  YAML flow lists split correctly, and a non-UTF-8 file fails with a
  message instead of a traceback. Module docstring lists check 4.
- CHANGELOG no longer claims "no new prompt": compile and aux cleanup,
  which the removed bare Bash covered in a skill-invoking turn, now
  prompt, deliberately. "Once per session" wording corrected to per call.
- Three upload-shaped deny siblings added (curl -XPOST, curl --request
  POST, wget --method=POST); /add-portal recon points at fetch_page.py.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>

.claude/commands/rank.md

@@ -58,7 +58,7 @@ Dispatch parallel `general-purpose` agents via the **Agent tool**, ~5 jobs per a
 - Agents fetch each posting URL following the escalation order in `.claude/skills/job-application-assistant/09-web-research.md` (using the installed portal CLI's `detail` command when the URL belongs to a portal in `.agents/skills/`, or `WebFetch` otherwise) and score **only from actually fetched content** - never from the title alone, never from fabricated posting content. A job that is not scored gets one of two statuses, and they mean different things:
   - `expired` - **positive evidence the posting is closed**: the page says it is no longer available or the deadline has passed, the server answers 410 Gone, or the URL redirects to a listing page instead of the posting.
   - `unreachable` - **retrieval failed, so nothing is known about the posting**: 401/403 or a login wall, a 404, a timeout, a 5xx, or an empty or unparseable page. The agent adds `"fetch_error"`, one short plain-text line naming the failure (e.g. `"403 after curl retry"`). A 404 is `unreachable`, not `expired`: links break while postings stay live (#538 was one).
-- **Before returning `unreachable` or `expired`, the agent must exhaust the escalation order** in `09-web-research.md`: installed portal CLI first, then `WebFetch`, then curl with browser headers (if `robots.txt` permits), then search for the employer's own careers site. A `WebFetch` 403 is a rejected *client*, not a missing page, and retrying with browser headers via curl recovers most corporate and bank domains. A stored URL ending in a `#fragment` points at a listing page rather than a posting, so the agent should search the employer's own careers site for the role by name before writing the job off. Include this instruction in every scoring agent's prompt. Neither status means "the first fetch was unhelpful".
+- **Before returning `unreachable` or `expired`, the agent must exhaust the escalation order** in `09-web-research.md`: installed portal CLI first, then `WebFetch`, then `python3 tools/fetch_page.py` with browser headers (it enforces `robots.txt` itself), then search for the employer's own careers site. A `WebFetch` 403 is a rejected *client*, not a missing page, and retrying with browser headers via `fetch_page.py` recovers most corporate and bank domains. A stored URL ending in a `#fragment` points at a listing page rather than a posting, so the agent should search the employer's own careers site for the role by name before writing the job off. Include this instruction in every scoring agent's prompt. Neither status means "the first fetch was unhelpful".
 - Scope is triage: posting text vs. rubric. **No company research, no salary lookup, no web searches** - that depth belongs to `/apply`.
 
 Each agent returns a JSON array, one object per job: