Repository navigation
Releases
v1.2.1
First release on PyPI, under the distribution name agent-perimeter-scan:
uv tool install agent-perimeter-scan. PyPI refused agent-perimeter as too similar to an
existing project. The command, the import package (agent_perimeter) and the Action are
unchanged; uses: Berakhah/agent-perimeter@v1 picks it up.
Changed
- The Python distribution is named
agent-perimeter-scan.agent_perimeter.__version__reads
the installed version under that name.
v1.2.0
First release on PyPI: uv tool install agent-perimeter. No change to action.yml's inputs or
outputs; uses: Berakhah/agent-perimeter@v1 picks it up.
Added
instrument: bring-your-own-agent injection proof. A loopback-only, scope-gated proxy
sits between your agent and a server, plants a per-run nonce canary in tool results, and
records a redacting transcript. A finding is rated by whether your agent repeated the canary
into a sink and what that sink can reach at call time. It refuses cross-origin, foreign-Host
and non-JSON requests. Seedocs/byo-agent.md.- Scan notes on the live-scan screen and in SARIF, including why a tool listing was incomplete.
examples/scan-reference-server.yml: a copy-paste workflow that scans the MCP reference
server (@modelcontextprotocol/server-everything@2.1.0) as a stdio target and uploads SARIF
to code scanning..github/workflows/examples.ymlruns it against every change..github/workflows/demo-gif.ymlrecordsdocs/demo/scan.giffromdocs/demo/scan.tapeon
a runner with Docker.- PyPI metadata (description, classifiers, project URLs) and a PyPI-specific description with
absolute links.
Fixed
- Tool listings fail closed.
tools/listfollowsnextCursorwith a page cap. A failed or
malformed first page, a repeated cursor or a page with notoolskey is an incomplete
listing: the scan says so, exits2on a gate run, and never writes or compares a drift
baseline from it. A server that returns zero tools is a complete, empty listing. The
server-supplied cursor is never echoed into notes. - The Action reports
drift: unknown, notfalse, when a gate run was refused before any
comparison. - Injection ratings: only sinks backed by an observed capability rate CRITICAL; a verbatim copy
of a tool result is LOW data pass-through; description-only sinks are named as such. - Instrument redaction: secrets glued to the canary, URL-encoded or nested several rounds deep,
and values under secret-sounding keys are redacted before anything is written. - Reports now carry the installed package version. SARIF
tool.driver.versionand the scan
record'stool_versionwere hardcoded to0.1.0; they readagent_perimeter.__version__,
which comes from the package metadata, and a test fails if a literal comes back. --htmlcrashed in any installed copy (the Action, and a PyPI install): importing the report
renderer readtests/fixtures/corpus.yaml, which only exists in a git checkout. The corpus is
now read on first use, and a test imports the CLI and renders a report from a copy of the
package with notests/beside it. Found by the newexamplesworkflow.- The eval never scores a fixture whose listing stopped early, and the degraded-mode floor test
can now fail.
Changed
- Security reports go through GitHub private vulnerability reporting; the personal email
address is gone and a test guards against a published one. The acknowledgement target is 5
business days. - Every third-party GitHub Action is pinned to a commit SHA.
- The source distribution contains the package, README, licence and NOTICE only (163 KB, down
from 2.8 MB); tests, fixtures, the web UI and census data stay in the git repo. - Dependencies: patched known vulnerabilities in the Python and web lockfiles (including Mako
1.4.3). - CI no longer uploads the deliberately flawed fixtures' SARIF to this repo's code scanning.
Removed
- Private project briefs, from the public repo.
v1.1.0
Changes since v1.0.0
- fix(census): Tier 3 (live-discover stratum) permanently dropped. Code review found
tier3.pysent live, unauthenticatedserver/discoverprobes to sampled third-party hosts with noScopeFilegate — a direct conflict with Never-rule 1 (no active probe without a scope file). Deleted, not redesigned: a seeded random sample of unowned servers has no single authorising party who could sign a scope file covering it. Census is passive-only (Tier 1 + Tier 2); seedocs/census/CHANGELOG.md. - feat(web): Data-Dense Ops Console (Serif Accent) redesign. New elevation/surface tokens, severity chip backgrounds, type scale and serif heading voice, motion presets (entrance/hover/focus states across scan-setup, live-scan, findings, capability graph, drift), a reusable
StatTileGrid, and aSparklineon the drift screen that never fabricates a flat line for single-scan history. - docs: drift-detection design spec marked implemented; ops-console redesign spec marked implemented; census report template and docs updated for the Tier 3 removal.
- No changes to
action.yml's inputs/outputs — this is a backward-compatible release.
Verified before tagging: full pytest suite (794 passed, 95% coverage), ruff + mypy --strict clean, Playwright suite (97 passed, incl. axe a11y and keyboard-only specs), docker compose up clean build/healthy/teardown, containerization/sandbox tests against a live Docker daemon (14 passed).
v1.0.0 - GitHub Action
First tagged release. uses: Berakhah/agent-perimeter@v1 - see README "Use as a GitHub Action".