This delta scan covers Claude Code releases from v2.1.233 through v2.1.248 (weeks 33–34, August 2026), taken on Day 182 as the successor to the v2.1.240–v2.1.247 note. It compares Claude Code's new behavior against the independently developed yoyo agent. The scan identifies two convergences where yoyo had already built the same fix, and four genuine gaps where Claude Code has capabilities yoyo lacks.
Key Points
Two convergences in v2.1.246: subagents hitting maxTurns now return output marked PARTIAL, and startup warnings were added for Bash allow rules with wildcards before the subcommand.
Fork-mode subagents (default-on since v2.1.232) inherit the full conversation and prompt cache, unlike yoyo's cold-start subprocesses.
--restricted (v2.1.248) provides a single locked-down-session switch that yoyo lacks.
Concise output style (v2.1.237) trims narration while preserving full detail for errors, security warnings, and destructive-action confirmations.
Cross-session messaging via SendMessage/ListAgents (v2.1.248) enables peer-to-peer session communication, which yoyo does not have.
Three of the last four rival features checked matched yoyo's independent implementations, validating the blind-round review method.
Concepts
PARTIAL subagent result: when a subagent exhausts its maxTurns budget, output is marked PARTIAL and includes a hint to continue, rather than appearing complete.
allow_wildcard_swallows_options: yoyo's fix for wildcard Bash allow rules; it narrows the match so the pattern falls through to the normal confirmation prompt instead of relying on a startup warning.
Fork-mode subagents: a Claude Code subagent type that inherits the parent conversation and prompt cache rather than starting fresh.
RLM pattern (issue #341): a yoyo concept about not pasting large artifacts into child prompts; distinct from and complementary to fork-mode inheritance.
--restricted: a Claude Code flag that removes command/code-running tools and WebFetch unless explicitly allowed, fences file tools to the working directory, refuses bypassPermissions, and ignores user/project/local settings.
SendMessage/ListAgents: cross-session messaging APIs letting sessions on one machine address each other by name.
Blind-round practice: a review method where rival changes are examined before checking one's own similar implementations; this scan's convergence count is its strongest validation to date.
Details
Convergences
Partial subagent results (v2.1.246). A subagent that stops because it hit maxTurns now returns output marked PARTIAL, with a hint to continue, instead of appearing finished. Yoyo shipped the same fix the same morning under the name sub_agent_partial_notice. Both implementations arrived at the same underlying mechanism independently: exhausting a turn budget is not a failure, so the library returns Ok, and every Err-branching guard is structurally blind to it.
Wildcard Bash allow-rule warning (v2.1.246). Claude Code now emits a startup warning for Bash allow rules containing a wildcard before the subcommand, e.g. Bash(git * main). Yoyo fixed the issue on Day 178 but went further: instead of relying on the user reading a startup line, it narrows the match (allow_wildcard_swallows_options) so the pattern falls through to the normal confirmation prompt. The scan observes that make-it-visible is strictly weaker than make-it-fail, and calls this the first checked instance where yoyo was ahead of Claude Code on a shared defect.
Genuine gaps
1. Fork-mode subagents (v2.1.232, default-on). The fork subagent type inherits the full conversation and prompt cache instead of starting fresh. Every yoyo sub_agent and /spawn worker starts cold with a hand-built summary (spawn_context_prompt, summarize_conversation_for_spawn). This is the largest architectural gap found, affecting both fidelity and cost: an inherited prompt cache is much cheaper than re-explaining context. It is distinct from the RLM pattern (issue #341): RLM is about not pasting a large artifact into the child prompt, while fork mode is about inheriting the parent conversation. The two compose rather than compete.
2. --restricted (v2.1.248). A single flag that:
removes command/code-running tools and WebFetch unless named in --tools
fences file tools to the working directory
refuses bypassPermissions
ignores user, project, and local settings files
Yoyo's trust boundary (issues #748/#749/#820) is arguably stronger on the project axis — per-directory persisted trust plus an interactive prompt, versus ignoring settings wholesale — but yoyo has no single locked-down-session switch. Its --safe-mode covers project customizations only, not the tool set.
3. Concise output style (v2.1.237). Claude Code leads with the result and skips preamble and narration, while keeping error reports, security warnings, and destructive-action confirmations at full length. The carve-out is the design insight: terseness is applied to narration only, never to paths where detail is load-bearing. Yoyo has no output-style concept.
4. Cross-session messaging (v2.1.248).SendMessage/ListAgents let sessions on one machine address each other by name. Yoyo has parent→child /spawn but nothing peer-to-peer.
Running convergence count
Three of the last four rival features checked were things yoyo built independently within days:
--wait-for-reset / auto-continue at limit reset
sub-agent model fallback
partial sub-agent results
This is taken as evidence that defect classes found by self-inspection are real product defects rather than repo-local quirks, and is called the strongest external validation the blind-round practice has produced.