clinical-trial-simulation

Name: clinical-trial-simulation
Author: RConsortium

88stars20forks

How to install clinical-trial-simulation?

rconsortium/pharma-skills/clinical-trial-simulation

$npx -y skills add RConsortium/pharma-skills --skill clinical-trial-simulation

Installs into the current project.

Summary

Design and simulate clinical trials using the TrialSimulator R package and produce a QC-ready build-order-spine report that pairs each block of code with rationale, parameters, and operating characteristics.

Files · 1

View on GitHub

SKILL.md

1	# TrialSimulator Skill
2
3	Help a biostatistician design and simulate a clinical trial using the
4	TrialSimulator R package, then write a readable report. This skill is
5	a thinking framework, a cached API reference, a TrialSimulator-specific
6	function catalog, and a report-writing guide. It is not a script.
7	You bring general engineering, programming, and biostatistics
8	knowledge; this skill adds what is specific to TrialSimulator.
9
10	## Loading announcement
11
12	Whenever this skill is loaded, the agent calls
13	`packageVersion("TrialSimulator")` (fast, local — no network) and
14	compares the installed version against
15	`metadata.trialsimulator_min_version` from the YAML frontmatter
16	above. The first line of the agent's first response reports the
17	result.
18
19	Three outcomes:
20
21	- Installed ≥ required — `clinical-trial-simulation v<metadata.version> loaded — TrialSimulator <ts_version>, R <r_version>`. Proceed.
22	- Installed < required — `clinical-trial-simulation v<metadata.version> loaded — TrialSimulator <ts_version> installed, ≥ <trialsimulator_min_version> required. R <r_version>.` Then offer to update on the next turn: "Want me to run `remotes::install_github('zhangh12/TrialSimulator')` now? (y/n)" — on confirm, run it and re-check; on decline, stop until the user resolves. Do not begin any simulation work until the version requirement is met.
23	- Not installed — `clinical-trial-simulation v<metadata.version> loaded — TrialSimulator not installed. R <r_version>.` Offer to install: "Want me to run `remotes::install_github('zhangh12/TrialSimulator')` now? (y/n)"
24
25	Substitutions:
26
27	- `<metadata.version>` from this file's YAML frontmatter.
28	- `<trialsimulator_min_version>` from this file's YAML frontmatter.
29	- `<ts_version>` from `packageVersion("TrialSimulator")`.
30	- `<r_version>` from `R.version.string` (the `x.y.z` portion).
31
32	No network call fires on load. The only network use is the optional
33	remediation install on user confirmation.
34
35	After the announcement line, briefly greet the user and ask about
36	the trial setting — never ask what design type, discover it through
37	conversation (see "Two user modes" below).
38
39	## Files in this skill
40
41	- `SKILL.md` (this file) — framework, conversation principles, build order, workflow
42	- `references/building_blocks.md` — cached reference for `endpoint`, `arm`, `trial`, `milestone`, `listener`, `controller`, `regimen`, and the condition system
43	- `references/helpers.md` — catalog of TrialSimulator-provided functions (RNGs, parameter solvers, analysis wrappers, post-sim utilities), plus non-obvious gotchas
44	- `references/report.md` — how to write the simulation report (intentionally policy-light; organizations are encouraged to edit this file)
45
46	These files cache the most common things to save tokens. When confused
47	or when behavior contradicts these notes, consult `?<function>` in R
48	or the package's pkgdown site at
49	https://zhangh12.github.io/TrialSimulator/. Don't guess — the manual
50	is the source of truth.
51
52	## Package source
53
54	This skill targets a known-good baseline of TrialSimulator declared
55	in `metadata.trialsimulator_min_version` of the YAML frontmatter
56	above. **At load, only a local `packageVersion("TrialSimulator")`
57	check runs — no automatic `install_github`, no network call.** The
58	load-time check + offer/auto behavior is specified in "Loading
59	announcement".
60
61	If TrialSimulator is missing or below the minimum, the canonical
62	remediation is:
63
64	```r
65	remotes::install_github("zhangh12/TrialSimulator")
66	```
67
68	The agent offers to run this when the load-time check detects a
69	shortfall and only runs it on user confirmation. The user can also
70	update outside the session and reload the skill.
71
72	When bumping `metadata.version`, audit whether
73	`trialsimulator_min_version` also needs to move — e.g., if the
74	skill adopts a TS feature or fix landed in a newer release.
75
76	Surface three versions in §0 of the report: TrialSimulator, R, and
77	the skill. The agent looks each value up once before writing the
78	report and pastes the literal string into the §0 table. No R chunks,
79	no runtime lookup, no `readRDS`. Slight staleness on re-renders is
80	acceptable; the running environment is assumed stable.
81
82	## Package philosophy
83
84	TrialSimulator decouples a trial into a small set of independent
85	building blocks: endpoints, arms, the trial object, milestones, the
86	listener, the controller, and (optionally) regimens for treatment
87	switching. **A trial design — fixed, seamless, response-adaptive,
88	dose-ranging, platform, anything — is just a particular composition of
89	these blocks.** There is no "design type" object; there are only blocks
90	and how they combine.
91
92	Practical implication: do not pick a design temp

People who liked this also viewed

6 picks

Type

clinical-trial-simulation

88stars20forks

View on GitHub

How to install clinical-trial-simulation?

rconsortium/pharma-skills/clinical-trial-simulation

$npx -y skills add RConsortium/pharma-skills --skill clinical-trial-simulation

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# TrialSimulator Skill
2
3	Help a biostatistician design and simulate a clinical trial using the
4	TrialSimulator R package, then write a readable report. This skill is
5	a thinking framework, a cached API reference, a TrialSimulator-specific
6	function catalog, and a report-writing guide. It is not a script.
7	You bring general engineering, programming, and biostatistics
8	knowledge; this skill adds what is specific to TrialSimulator.
9
10	## Loading announcement
11
12	Whenever this skill is loaded, the agent calls
13	`packageVersion("TrialSimulator")` (fast, local — no network) and
14	compares the installed version against
15	`metadata.trialsimulator_min_version` from the YAML frontmatter
16	above. The first line of the agent's first response reports the
17	result.
18
19	Three outcomes:
20
21	- Installed ≥ required — `clinical-trial-simulation v<metadata.version> loaded — TrialSimulator <ts_version>, R <r_version>`. Proceed.
22	- Installed < required — `clinical-trial-simulation v<metadata.version> loaded — TrialSimulator <ts_version> installed, ≥ <trialsimulator_min_version> required. R <r_version>.` Then offer to update on the next turn: "Want me to run `remotes::install_github('zhangh12/TrialSimulator')` now? (y/n)" — on confirm, run it and re-check; on decline, stop until the user resolves. Do not begin any simulation work until the version requirement is met.
23	- Not installed — `clinical-trial-simulation v<metadata.version> loaded — TrialSimulator not installed. R <r_version>.` Offer to install: "Want me to run `remotes::install_github('zhangh12/TrialSimulator')` now? (y/n)"
24
25	Substitutions:
26
27	- `<metadata.version>` from this file's YAML frontmatter.
28	- `<trialsimulator_min_version>` from this file's YAML frontmatter.
29	- `<ts_version>` from `packageVersion("TrialSimulator")`.
30	- `<r_version>` from `R.version.string` (the `x.y.z` portion).
31
32	No network call fires on load. The only network use is the optional
33	remediation install on user confirmation.
34
35	After the announcement line, briefly greet the user and ask about
36	the trial setting — never ask what design type, discover it through
37	conversation (see "Two user modes" below).
38
39	## Files in this skill
40
41	- `SKILL.md` (this file) — framework, conversation principles, build order, workflow
42	- `references/building_blocks.md` — cached reference for `endpoint`, `arm`, `trial`, `milestone`, `listener`, `controller`, `regimen`, and the condition system
43	- `references/helpers.md` — catalog of TrialSimulator-provided functions (RNGs, parameter solvers, analysis wrappers, post-sim utilities), plus non-obvious gotchas
44	- `references/report.md` — how to write the simulation report (intentionally policy-light; organizations are encouraged to edit this file)
45
46	These files cache the most common things to save tokens. When confused
47	or when behavior contradicts these notes, consult `?<function>` in R
48	or the package's pkgdown site at
49	https://zhangh12.github.io/TrialSimulator/. Don't guess — the manual
50	is the source of truth.
51
52	## Package source
53
54	This skill targets a known-good baseline of TrialSimulator declared
55	in `metadata.trialsimulator_min_version` of the YAML frontmatter
56	above. **At load, only a local `packageVersion("TrialSimulator")`
57	check runs — no automatic `install_github`, no network call.** The
58	load-time check + offer/auto behavior is specified in "Loading
59	announcement".
60
61	If TrialSimulator is missing or below the minimum, the canonical
62	remediation is:
63
64	```r
65	remotes::install_github("zhangh12/TrialSimulator")
66	```
67
68	The agent offers to run this when the load-time check detects a
69	shortfall and only runs it on user confirmation. The user can also
70	update outside the session and reload the skill.
71
72	When bumping `metadata.version`, audit whether
73	`trialsimulator_min_version` also needs to move — e.g., if the
74	skill adopts a TS feature or fix landed in a newer release.
75
76	Surface three versions in §0 of the report: TrialSimulator, R, and
77	the skill. The agent looks each value up once before writing the
78	report and pastes the literal string into the §0 table. No R chunks,
79	no runtime lookup, no `readRDS`. Slight staleness on re-renders is
80	acceptable; the running environment is assumed stable.
81
82	## Package philosophy
83
84	TrialSimulator decouples a trial into a small set of independent
85	building blocks: endpoints, arms, the trial object, milestones, the
86	listener, the controller, and (optionally) regimens for treatment
87	switching. **A trial design — fixed, seamless, response-adaptive,
88	dose-ranging, platform, anything — is just a particular composition of
89	these blocks.** There is no "design type" object; there are only blocks
90	and how they combine.
91
92	Practical implication: do not pick a design temp

People who liked this also viewed

6 picks

Type