Install¶
Install the skill with the standard Skills installer (Node.js/npm required); no developer clone or development dependency sync is needed:
Use Python 3.11+ and uv. Resolve the actual directory returned by the installer; the shared default is normally ~/.agents/skills/acquire-research-papers:
$skill = Join-Path $env:USERPROFILE ".agents/skills/acquire-research-papers"
python "$skill/scripts/arp.py" --help
python "$skill/scripts/arp.py" config status
The standard-library launcher reads runtime.json, downloads the fixed release wheel, verifies SHA-256, and installs it with uv into a separate cache. It never needs a developer checkout or installs development tools into the skill. File configuration and core commands support Windows/macOS/Linux. Current IEEE browser access and legacy DPAPI require Windows, PowerShell, and Node.js; on macOS and Linux, IEEE papers join the manual-download queue. Live institutional access needs a separate check.
Update by rerunning the same Skills install command. Registries, credentials, caches, named institution profiles, and browser sessions live outside the installed skill. Use a separate delivery directory. Development builds can use the digest-verified local-wheel procedure in CONTRIBUTING.md.
Developer checkouts
If your current skill directory is a Git developer checkout, preserve it. Do not point a third-party installer's overwrite operation at that directory. Build an export and use the repository's safe copy installer into a new directory; it refuses .git targets and keeps an explicit backup when replacing a clean previous thin skill. See safe migration.
Install only the CLI¶
From v0.16.0 the CLI wheel is also on PyPI, uploaded by the release provenance workflow after it has verified the wheel against the published release (see Verify a release):
This installs the arp command without the agent skill. Agents still need the skill installed as above; its launcher keeps using its own pinned and digest-checked wheel.
Host setup and checked support¶
Codex, Claude Code, Kiro IDE, and Kiro CLI share the same skill and pinned wheel. User skill directories are ~/.agents/skills/, ~/.claude/skills/, and ~/.kiro/skills/ respectively. Copy only the exported acquire-research-papers/ folder. Kiro custom agents require explicit skill:// resources; a resource template is included. See host installation and official sources, checked on October 20, 2026.
Run arp doctor --agent-host <codex|claude-code|kiro-ide|kiro-cli> --skill-directory <actual-path> (--host is the same option; Kiro's shell refuses command lines that contain --host). Its default local inventory reports Python/runtime/schema/configuration/dependency state. Host loading and institutional access remain unverified. --check-host-version explicitly runs the host executable's version command; --remote-check explicitly probes public Crossref reachability without credentials. Official format compatibility and synthetic CI are not live certification; Claude Code and Kiro interactive execution remain unverified until recorded.
Release assets and manual installation¶
Every GitHub release provides:
| Asset | Use |
|---|---|
acquire-research-papers-skill-<version>.zip |
Extract and copy the contained acquire-research-papers/ folder to a new skill directory. It contains instructions, references, schemas, a launcher, and the pinned runtime manifest. |
acquire_research_papers-<version>-py3-none-any.whl |
Install the CLI independently with uv tool install <wheel-path>, or supply its absolute path as ARP_LOCAL_WHEEL when bootstrapping offline. The launcher still verifies the manifest digest. |
SHA256SUMS |
Compare downloaded asset hashes with Get-FileHash <path> -Algorithm SHA256 on PowerShell or sha256sum <path> on Linux. |
Manual skill installation needs Python 3.11+ and uv. The launcher normally downloads its pinned wheel on first use; dependency installation also needs network access or a populated uv cache. Node.js and Google Chrome are additionally required for IEEE browser access. Do not copy the source archive or a developer checkout into the skill directory. See SUPPORT.md for bootstrap and migration troubleshooting.
Verify a release¶
From v0.16.0 on, the release provenance workflow rebuilds every published release from its tag on Linux and on Windows. It requires both rebuilds, the tag's runtime.json pin, and the published wheel and ZIP to be identical byte for byte, and then records a GitHub artifact attestation for the two files. Check a downloaded asset with the GitHub CLI:
gh attestation verify acquire_research_papers-<version>-py3-none-any.whl --repo EnosElinsa/acquire-research-papers
gh attestation verify acquire-research-papers-skill-<version>.zip --repo EnosElinsa/acquire-research-papers
A passing check means the file's SHA-256 matches a build that this repository's workflow produced from the release tag. Releases before v0.16.0 have no attestation; compare their hashes with SHA256SUMS instead.