Add the environments reference to the skill and install it with setup: record
against a disposable instance, sign in through the UI, write with request
steps and read each write back, check coverage, re-record by withheld reason
and export the bundle. Route it from SKILL.md, name agent environments in the
skill description, and document the request step, the export tool and
recording writes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
export_environment_bundle writes the bundle environment-build writes, from
capture Evidence retained in the session or one capture Evidence file,
into an absolute output directory that must not exist and whose parent
must exist, as extract_firmware and export_web_scripts require; it never
creates the parent or reuses a directory. It returns the same report,
whose schema now lives in the domain so the CLI and the tool share it, and
dry_run reports without writing. A loopback capture without a file needs a
name. The build's refusal messages no longer name CLI flags.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
verify:browser:request records a write session against a fixture item API
that refuses writes without its CSRF cookie header and stored bearer
token. It checks that every request lands in its own step, that the page
values authenticated each write and appear in no record, header or body,
that an unexpected status fails the step and cancels the rest, and that
environment-build serves the recorded create, update and delete.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A capture_browser_scenario step `request` sends one declared HTTP request
through the page's own fetch, with its cookies and session, the way the
application's frontend would, and reads the whole response so the capture
records the transaction like any other. The path must resolve to the start
URL's origin, and the page must be on that origin when the step runs.
Query values and header values may be literals or declared secrets, and a
header may read its value from a cookie or a local storage entry, which is
then redacted from later records like a declared secret. `expected_status`
fails the step on any other status. Scenarios without a request step are
unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The manifest, state and coverage stages run synchronously, so a SIGINT
that arrived during them was only dispatched once the bundle write had
begun. Each check now yields first, so an interrupt before the bundle
stage cancels the build with nothing written.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The guides now state when the one-step bundle equals the three-step one,
which headers inspection excludes, what the credential warning scans, how
cancellation behaves, and that the coverage view does not apply the
serve-time admission checks. The spec folder records the census, the end
to end runs, the gates and the review.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The credential scan walks with its own stack, so a deeply nested body
cannot crash the build, and it also reports credential-named header
names, query names, form body members and string arrays.
- The build checks SIGINT between stages: cancelled before the bundle
stage it writes nothing, during the write it names the written bundle.
- A dry run opens the package tarball, so it refuses what a build refuses.
- Repeated template rules are refused, a HAR saved as .json gets a hint,
file-derived names follow the bundle name rules, `localhost.` counts as
a loopback origin, and the next command is shell-quoted.
- --capture-timeout-ms passes through to the HAR inspection, and a
missing directory is refused before inspection.
- The no-observation hint covers updates and deletes of entities the state
never held; the coverage view states that it does not apply the
serve-time admission checks; the inspect_analysis_view description
names the view.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The agent environments guide opens with the one-step build: the report,
the hint for each withheld write reason, and what a bundle ships. The CLI
guide, the capture guide and the skill reference name the command and the
environment-coverage view.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The lane now runs the command users run: it builds the Grafana-shaped HAR
with environment-build and the packed checkout, checks the report, then
runs the bundle with agent-env as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Domain tests check the hint table against the state view's reasons, the
summary, and the view against the environment-state view it summarizes.
Service tests build the fixture capture and compare every file with the
three-step path, a dry run, loopback naming and the refusals. A CLI test
compares a HAR built in one step with the three CLI steps byte for byte,
and an MCP test projects the view from every source kind.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
rea environment-build <capture> [directory] inspects a HAR (or reads
capture Evidence), projects the environment manifest from the in-memory
Evidence, authenticates the pair, prepares the state and writes the bundle
environment-export writes. Every bundle file is byte-identical to the
three-step path for the same capture and options.
The report adds the capture and manifest identities, the coverage summary,
warnings (a loopback capture named after its file, credential-named body
members by pointer) and the next command. --dry-run writes nothing. The
default template rules are the four the published environments use.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A new inspect_analysis_view kind over capture Evidence. It projects the
state an environment derives with the given template rules and reports
each tool as answered from state, replayed with its view reason, a served
write, or a withheld write with its reason, the records behind it and a
hint about what to record next. The hints cover every write reason, so a
new reason without one fails to compile.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Resolve application and managed workflow contracts through the canonical name-based resolver, removing catalog-position overloads and the obsolete local facade.
Use the existing retained-summary policy for eight offline registrations. Pass only validated delivery detail to that owner while preserving exact typed service inputs and validation, admission, logging, and Evidence recording order.
Align export-shape MCP parity coverage with the existing absolute-path contract using an absolute dot alias. Preserve exact caller-path parameters, canonical artifact parity, distinct Evidence identity, and the export comparison assertions.
Resolve declaring TypeDef tokens within their own source and reuse the retained nominal identities for pairing, absence, and aggregate coverage. Remove display-name membership interpretation while preserving partial known pairs and whole-inventory matching.
A generated managed fixture reproduced false additions and removals for distinct nested owners sharing a display name before this change. All 305 focused producer and public CLI-to-stdio-MCP checks passed, including comparisons after source removal, missing identities, and genuine whole-inventory changes. Clarify selected-scope semantics in the docs; no live managed runtime was used.
Select registration without skill installation in the synthetic-client recovery matrix. Preserve all four specific remediation assertions and verify the plan, failure reason, and absence of unrelated installation work.
Route MCP snapshot reads through the session's existing file-resource owner so cleanup remains visible, blocks replacement, and can be retried by close_binary. Keep reads inside the serialized open callback and remove superseded adapter resource plumbing.
Validate typed errors, exact descriptor recovery, unchanged source bytes, replay provenance, and same-connection lifecycle ordering. Focused snapshot, target cleanup, and MCP lifecycle checks passed with simulated providers; no live provider was exercised.
Fixes#6216
Classify selected-input read failures at the shared artifact boundary and remove duplicate phase-based translations from TShark and sigrok. Keep other read stages raw for their existing owners and retain cleanup resources until retry succeeds.
Extend the descriptor outcome matrix with post-run private-copy integrity failures, late cancellation, retained command output, failed cleanup, and successful retry.
Project admitted output facts beside the command owner and remove adapter-local copies. Preserve byte-count omission, explicit signal and exit facts, and caller-specific truncation and failure policies.
Update ADB timeout regressions to assert the existing prerequisite frame and original process cause, including captured output on both frames.
Derive aggregate selected-scope agreement from retained nominal type identities instead of requested spellings. Keep incomplete or historical selected scope unknown while preserving caller selections, whole-inventory distinction, paging and member absence behavior.
Cover equivalent requests, same-label nested owners, missing and partial identity facts, and complete empty selections. Exercise CLI inspection Evidence through stdio MCP comparison after removing its selected source file.
Keep the selected running JVM visible when its version is unsupported, while retaining missing-runtime fallback controls. Assert the native snapshot binding and composed workflow Evidence profiles separately, preserving their exact commitments through a rejected snapshot switch.
Preserve source code-point prefix endpoints in the shared marshal UTF-8 decoder, including separately encoded surrogate values. Keep native TextDecoder spans, full validation, byte identity and the existing 80-unit budget. Update the PyInstaller consumer without changing its text contract.
Validation: 207 relevant source tests;24 affected compiled CLI/MCP/filesystem CI tests; independent CPython3.11.2 compiler/marshal oracle and real argparse;18 equivalent performance tasks; build/check passed. Codex reviewed7c0f7abc6b without major findings.
CI bypass authorized for demonstrated baseline-only failures:101 source and179 built cases reproduced on actual tested main ba9c5bc68. Cancelled shard remainder is unverified. See PR body and discussion for scope and evidence.
Fixes#6476
Parse declarations before selecting values, admit safe integers once, and retain bounded raw numeric issues. Preserve display and package selection policies while distinguishing absent and malformed values.
Keep completed display facts in typed physical-size failures, share the existing diagnostic sample limit, and verify real CLI/MCP contracts with a strict fake ADB.
Parse MIME parameters once with strict quoted-value and duplicate handling,
and compare equivalent default binary profiles. Keep excluded HAR MIME
unknown, explicit framing independent, and nominal gRPC sanitization intact
across malformed or repeated declarations.
Add public CLI/MCP and native-header regressions with source identity and
partial-coverage assertions. Preserve upstream wire and schema admission.
Fixes#6836
Stop interactions and sampling before final observation, and stop the owned
target before failure-path rendering. Cache finalized terminal facts through
disposal and cleanup retry while preserving the original failure.
Version deferred terminal observation order for every normalization profile,
including disabled PID and port normalization. Add public CLI/MCP descendant,
partial-evidence and compatibility controls without replacing upstream owners.
Fixes#6443
Share parser-owned package and runtime facts across dependency consumers.
Preserve uncertain loader requests and explicit wrapper bindings without
inventing unbound require authority in known modules. Cover selected source
grammar through the public MCP boundary and retain ambient/type dependencies.
Open each selected directory before loading its scoped policy, retaining
partial directory failures and cleanup ownership. Preserve explicit caller
negation precedence and source diagnostics across inventory projections.
Clarify how version comparison retains container provenance without claiming
semantic drift. Use independent unknown source inputs to keep candidate-limit
fixtures meaningful as conditional branch evaluation becomes more precise.
Fixes#6891Fixes#6882Fixes#6883Fixes#6859
Validate the selected workspace lookup through stat before canonicalization,
preserving typed configuration causes and physical traversal semantics. Cover
readiness without writes and tie startup cleanup regressions to explicit
cancellation or timeout producer outcomes.
Fixes#6828
Omit absent inventory facts without erasing explicit unknowns. Compare canonical file occurrences by source path and digest while preserving complete container provenance in observations and Evidence.
Consolidate candidate-capacity fixtures around unresolved alternatives and update inventory coverage and source-occurrence oracles. Preserve resource, identity, relationship uncertainty, cancellation, and transfer regressions.
* fix(pe): accept WIN_CERTIFICATE padding counted in dwLength
* perf(pe): reuse decoded PKCS#7 ASN.1 tree
* style: format files inherited from main for CI
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(ghidra): report kernel OOM kills as a memory constraint
A kernel or container OOM killer ends Ghidra with SIGKILL and prints no
java.lang.OutOfMemoryError, so the failure fell through to the generic
execution_failure that points at `rea doctor`.
On Linux with cgroup v2, read the process's own memory.events oom_kill
count before launch and again after a SIGKILL. When it rose, project
resource_constraint (memory) with the group's memory.max, and advise on
the whole process tree (native decompiler processes included) instead of
a larger Java heap. Otherwise name the signal as a possible OOM kill.
Fixes#5859
* fix(ghidra): scope OOM-kill attribution to what the cgroup proves
Review follow-up for the SIGKILL projection:
- The process supervisor records stop_requested before it sends any
signal, and Ghidra diagnostics carry it, so a SIGKILL from REA's own
stop is never reported as an outside kill. Only process and protocol
failures qualify; launcher (start) failures can include REA's rollback.
- Word the memory constraint as an OOM kill in REA's cgroup or a child
group; the count is per cgroup, not per process tree.
- Keep the open_binary reopen guidance for an established session.
- Resolve the cgroup directory through the cgroup2 mount in
/proc/self/mountinfo, so subtree mounts are joined from their root,
and read kernel files to EOF within a budget.
- Report memory_max as "max" for an unlimited group and null only when
the limit could not be read.
The showcase illustration was a single light PNG, so it stayed a white
slab on GitHub's dark themes. Redraw it as a light and a dark SVG, and
swap the linked image in README.md and all 21 translations for a
<picture> with a prefers-color-scheme: dark source. Alt text and link
targets are unchanged. The PNG is deleted because nothing else uses it.
website/figures.md notes the pair.
🤖 Generated with Claude Code (Sonnet 5.5) 🦊
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
