Glossary¶
Terms used across these docs, in alphabetical order. Each entry links to the page that explains it.
Actionable finding¶
An unsuppressed finding at or above the severity threshold that applies to its scanner. Actionable findings are what fail_on_findings counts: any actionable finding makes the scan exit 2. See Set the gate.
Age bound¶
The age past which ASH treats a content database as too old to trust: 120 hours for grype's database, 24 hours for trivy's, and 30 days for the offline rulesets. See Content databases.
Aggregated results¶
ash_aggregated_results.json, the file ASH writes at the root of the output directory. It holds the resolved configuration, with each value read from the environment recorded as its reference, a status row per scanner and converter, and the aggregated SARIF, and every reporter renders it. See How ASH works.
Aggregated SARIF¶
The single SARIF run that every scanner's results are merged into, written as reports/ash.sarif. Findings are appended, not deduplicated. See How ASH works.
AWS plugins¶
The ash_aws_plugins package, which ships with ASH and holds reporters that send results to AWS services. It loads only when the operator names it in a plugin module list; a repository config cannot. See AWS reporters.
Client-supplied¶
The mark ASH puts on a suppression or ignore path set by a config file an MCP client delivered, rather than by the operator. A client-supplied ignore path keeps the finding in the results as a suppressed finding instead of dropping it. See Configuration trust model.
Community plugin¶
A plugin package that ships with ASH but is not built in, such as the trivy-repo, snyk-code, and ferret-scan scanners. ash scan loads it only when a plugin module list names it; for the MCP server, see What a plugin can do. See Community plugins.
Content database¶
Data a scanner downloads and matches code against: grype's and trivy's vulnerability databases, and the rulesets semgrep and opengrep use offline. Each has an age bound, and by default a scan fails when one is older than its bound. See Content databases.
Converter¶
A plugin that runs in the convert phase and turns files scanners cannot read, such as archives and Jupyter notebooks, into files they can. Its output is scanned as the converted target. See How ASH works and the converter reference.
Delivered tree¶
Source code an MCP client sent to an ASH MCP server, as a git URL or a zip upload. The server keeps it under the MCP workspace root, and its config files belong to the MCP-client tier. See Run the MCP server.
Event subscriber¶
A plugin that runs code when a scan event happens, such as a phase starting or a scanner finishing. See Subscribe to scan events.
Exit code¶
The number ash scan returns: 0 for success, 1 for an incomplete or failed scan, 2 for actionable findings, 3 for a configuration it cannot use or an input it refuses before scanning, such as a linked output directory, 4 for a workspace error, and 5 for a usage error, such as an unknown option. See Exit codes.
Finding¶
One result a scanner reported, which is one result in the SARIF. Two scanners reporting the same problem produce two findings.
Gate¶
The rules that turn a scan's results into its exit code: fail_on_incomplete_scanners, content_db_staleness, fail_on_findings, and the severity thresholds. See Set the gate.
Ignore path¶
An entry in global_settings.ignore_paths. Findings located in matching files are dropped from the results. When it has an expiration date, it stops applying on that date, as a suppression does. See Suppressions, ignore paths, and scanner exclusions.
In-tree tool¶
A program found inside the scanned directory, for example in an in-repo .venv/bin or node_modules/.bin. ASH does not run in-tree tools unless the operator passes --allow-in-tree-tools, sets ASH_ALLOW_IN_TREE_TOOLS, or sets allow_in_tree_tools in the --cli-json-input document. ash mcp never runs them. See Manage scanner tools.
Incomplete scan¶
A scan in which a selected scanner is MISSING or ERROR, no scanner ran, or a converter could not convert all of its inputs. With the default fail_on_incomplete_scanners, an incomplete scan exits 1 whatever it found. See Set the gate.
Inline suppression¶
A comment in the source that suppresses a finding on its line: ASH's ash-ignore and ash-ignore-next-line, or a scanner's own syntax such as bandit's # nosec. See Suppressions, ignore paths, and scanner exclusions.
MCP client¶
A program that uses the ASH MCP server, such as an AI coding agent or an IDE. Files it delivers to the server form the MCP-client trust tier. See Run the MCP server and Configuration trust model.
MCP workspace root¶
The directory where the ASH MCP server keeps each session's delivered source and config: ASH_MCP_WORKSPACE_ROOT, else $XDG_CACHE_HOME/ash-mcp, else ~/.cache/ash-mcp. See Configuration trust model.
MISSING¶
The status of a scanner that was selected but did not run, because its tool was not found or was found only inside the scanned directory, or because it cannot run offline, as npm-audit cannot. The reason is in the scanner's missing_reason. A MISSING scanner makes the scan incomplete. See Troubleshoot.
Offline mode¶
A scan with --offline or ASH_OFFLINE set: every scanner runs in its offline mode, except npm-audit, which does not run offline, and ASH downloads and installs nothing. Content databases must already be in the cache. See Run ASH offline.
Operator¶
The person or pipeline running ASH. The operator owns the command line, the environment, and config files outside every git checkout and the scanned directory, and is the only tier that may grant sandbox access or choose where ASH's AWS reporters send results. See Configuration trust model.
Output directory¶
The directory ASH writes its results, reports, and logs to: --output-dir, else .ash/ash_output inside the scanned directory. ASH leaves it out of every scan, apart from the files git tracks in it. See Output files and How ASH works.
Phase¶
One stage of a scan: convert, scan, or report, run in that order, plus the optional inspect phase. See How ASH works.
Plugin module¶
A Python module that ASH imports to register plugins, named in ash_plugin_modules, --config-overrides, --ash-plugin-modules, ASH_PLUGIN_MODULES, or the Python API's ash_plugin_modules argument. It runs inside ASH's own process, outside the scanner sandbox. See Plugins.
Reporter¶
A plugin that renders the aggregated results in one format and writes it to the reports/ directory. See the reporter reference.
Repository config¶
An ASH config file inside a git checkout or inside the scanned directory, such as the .ash/.ash.yaml ASH discovers in the repository, or a config kept in an ops repository of its own. It cannot grant sandbox access, load ASH's off-host plugin packages, or choose where ASH's AWS reporters send results. See Configuration trust model.
Run mode¶
Where a scan runs: local (the default), container, nix, or precommit, chosen with --mode. See How ASH works.
Sandbox grant¶
A setting that gives sandboxed scanners more than the default policy: sandbox.network_scanners, sandbox.extra_read_paths, sandbox.read_path_scanners, or sandbox.env_scanners. Only the operator may grant. See Scanner sandbox.
Scanned directory¶
The directory a scan runs over: --source-dir, the working directory, or a workspace's root. A config file inside it is a repository config, and for the tool files you name ASH also looks at the scanned tree around it.
Scanned tree¶
The outermost git checkout at or above the scanned directory, or the scanned directory itself when it is not in a checkout. ASH treats every file inside it as written by the repository being scanned, so a tool file you name there is refused for the scanners that do not allow one. For every restricted setting a config file makes, a config file inside any git checkout, the scanned one or a separate one such as an ops repository, is a repository config. See Configuration trust model.
Scanner¶
A plugin that runs a security tool against a target and returns its results as SARIF. See the scanner reference.
Scanner exclusion¶
A scanner's own option that keeps files away from it, such as bandit's excluded_paths or checkov's skip_path. The scanner never sees the files, so nothing about them reaches the results. See Suppressions, ignore paths, and scanner exclusions.
Scanner sandbox¶
An operating-system confinement around the processes a scanner starts through ASH, with limited reads and writes, a filtered environment, and no network unless the scanner needs one; how strict it is depends on the backend. Off by default. See Scanner sandbox.
Scanner status¶
The result ASH records for each scanner and target. PASSED: it ran and found nothing actionable. FAILED: it ran and found actionable findings. ERROR: ASH tried to run it and it failed, or its tool could not start as asked. MISSING: it was selected but could not run. SKIPPED: it was not selected or was excluded, --python-only left it out, it does not support this platform, or it had nothing to scan. See How ASH works.
Severity threshold¶
The least severe level that counts as actionable: global_settings.severity_threshold, MEDIUM by default, or a scanner's own severity_threshold option, which replaces it for that scanner. See Set the gate.
Shard¶
One of several scans of the same repository that each run a share of the scanners, combined afterward with ash merge. See Scale to large repositories.
Suppression¶
An entry in global_settings.suppressions, or an ash-ignore comment, that keeps a finding in the results but marks it suppressed, with a reason. Suppressed findings do not count toward the gate. See Suppressions, ignore paths, and scanner exclusions.
Target¶
What a scanner scans: source, the scanned directory, or converted, the files the converters produced. See How ASH works.
Trust tier¶
The owner ASH assigns to a piece of configuration: operator, repository, or MCP client. The tier decides which settings the configuration may make. See Configuration trust model.
Workspace¶
A set of projects scanned together in one run, defined by a VS Code .code-workspace file, from the CLI or over MCP. See Scale to large repositories.