The JWT chain scanner only accepted a header at the start of a segment or
after `-`, so `session_<JWT>` reached the audit and cost logs unredacted.
Treat `_` as a prefix boundary too and cover it in the tests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Before 3.12 the parser rejects f"{d["k"]}" outright, so the shared sample
stopped at a parse error and the per-finding assertions failed on 3.11.
Samples that every supported Python parses now carry the per-finding
checks; the newer f-string forms run one per sample and accept either the
specific finding or the parse error.
Also flag unquoted X | Y annotations, which can fail at definition time
before 3.10, and check that a quoted union is not flagged.
The lookbehind added to keep the JWT scan linear also skipped tokens right
after a `-`, so `--cookie "session=jwt-<JWT>"` reached the logs unredacted.
Scan each `[A-Za-z0-9_.-]` run once instead, checking every dot segment for
a header at its start or after a `-`, decoding at most once per segment so
long `ey-ey-...` runs stay linear.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A later batch save advances evaluated_at past edits to previously evaluated skills. Compare each saved evaluation mtime to the current inventory, including timestamps that move backward, and retain the legacy global timestamp fallback.
ast.parse(feature_version=...) on current Pythons goes no lower than the
3.7 grammar, and it accepts f-string forms that older interpreters reject:
the self-documenting {x=} (3.8) and, from 3.12, reused quotes, backslashes
and comments inside a replacement field. ast does not record these, so
the guard now reads each f-string's own source text. The sample file
proves both kinds are flagged and that 3.6-valid fields ({n == 1},
{n!r:>4}, escaped {{n=}}) are not.
instinct-cli keeps running on old interpreters, but people on one should
know they no longer get security fixes. On a Python older than 3.11 (3.10
reached end of life in October 2026) the CLI now prints one line to
stderr after parsing its arguments:
[instinct-cli] Python 3.8 is past end of life. instinct-cli still runs
on it; upgrade to Python 3.11 or newer for security fixes.
The command itself runs as before. The notice stays out of automated
output: `prune --quiet`, which observer-loop.sh runs, skips it, and
CLV2_SKIP_PYTHON_EOL_NOTICE=1 turns it off for anyone who keeps an old
Python on purpose. The threshold is one constant, _SUPPORTED_PYTHON, to
raise as versions age out. Importing the module prints nothing.
The hooks fall back to any Python 3 that runs, so users on an older
interpreter (RHEL 8's 3.6, for one) still reach instinct-cli. It failed
there at import: builtin generics such as list[dict] in signatures are
evaluated when the function is defined, which raises TypeError before
3.9, and str.removeprefix is 3.9+.
- Quote the builtin-generic annotations. Type checkers read them the
same way, and they stay valid under 3.14's deferred annotations.
- Strip the byte-order mark with startswith and a slice.
- Open the registry lock file with an explicit UTF-8 encoding.
- Say in the module docstring that 3.6 is the floor and what keeps it.
The new test parses the script with the oldest grammar ast offers and
rejects what a grammar check cannot see: unquoted builtin generics in evaluated
annotations, `from __future__ import annotations`, and 3.7+ methods and
keywords (capture_output, text, removeprefix and similar). A sample file
proves it catches each of those. It also runs import, status, export,
evolve, projects and prune under -X dev -X warn_default_encoding
-W error, so any deprecation or default-encoding warning on the current
Python fails the build. CI runs only current Pythons; this is a static
guard, not a run on a real 3.6.
commands-core copies scripts/lib into the Claude config root, so it does not repair a marketplace plugin installation. Lead with a direct dashboard run from a full checkout and keep commands-core as the non-marketplace reinstall path.
Co-Authored-By: TRAE <noreply@trae.ai>
The diagnostic always blamed scripts/lib/skill-evolution, hiding a nested failure such as a missing lib/utils.js, and the printed installer command selected no module. Include the module from the resolution error and point at a full-checkout install with --modules commands-core.
Co-Authored-By: TRAE <noreply@trae.ai>
- Replace the secret=/api_key= value regex with a single-pass, quote-aware
scanner so the whole Bash word is redacted when quoted and unquoted
fragments are joined (e.g. --secret='prefix-'suffix), while a value inside
an enclosing quoted string stops at that string's closing quote.
- Searching for the keyword itself (grep -n "secret =" file,
rg 'api_key:' dir) no longer redacts the rest of the command.
- Use a lookbehind instead of \b before the JWT header so long dot-free
ey-ey-... runs are scanned in linear time instead of quadratic.
- Add regression tests for joined fragments, keyword searches, and the
ey- performance case.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A managed plugin install can lack scripts/lib/skill-evolution, which made skills-health fail on the first require with Cannot find module. Catch that specific resolution failure and print the install-apply remedy; other errors still propagate.
Co-Authored-By: TRAE <noreply@trae.ai>
Match JWT headers that start with ey or ew (whitespace after the opening
brace) and keep non-eyJ matches only when the header does not decode to
an object with an alg field. Parse secret= / api_key= values with escaped
double quotes, $'...' quoting and backslash escapes, and redact an
unterminated quoted value to the end of the command.
Refs #3846
sanitizeCommand only redacted keyword- or flag-shaped secrets, so a bare
Anthropic/OpenAI sk- key, a JWT, a PEM private key block, or a secret= /
api_key= / x-api-key: value was written verbatim to bash-commands.log and
cost-tracker.log. Add value-shape patterns for these and keep the key name
and separator for assignments so the log stays readable.
Refs #3846
On Windows, piped stdout defaults to cp1252, so `python3 gacha.py` crashed
on its first banner line with UnicodeEncodeError. That covers agent shell
tools and gacha.sh. Reconfigure stdout to UTF-8 before main(), the same
approach instinct-cli.py took for #1855.
The existing tests set PYTHONUTF8=1, which hid the crash. The new case
forces PYTHONIOENCODING=cp1252 without PYTHONUTF8, so it reproduces on any
OS: it fails on the previous gacha.py and passes with this change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Collect string-valued manifest.hooks as a file reference so missing files, directories, and paths outside the cache fail validation. Add regression coverage for valid and invalid hook references while preserving non-string and absent-hook behavior.
Closes#3786